diff --git a/examples/community/adaptive_mask_inpainting.py b/examples/community/adaptive_mask_inpainting.py
index da67debe72ec..a22ad0b54846 100644
--- a/examples/community/adaptive_mask_inpainting.py
+++ b/examples/community/adaptive_mask_inpainting.py
@@ -489,7 +489,6 @@ def enable_model_cpu_offload(self, gpu_id=0):
         # We'll offload the last model manually.
         self.final_offload_hook = hook
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -651,7 +650,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -666,7 +665,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/fresco_v2v.py b/examples/community/fresco_v2v.py
index b79834db5eed..abc5af8458af 100644
--- a/examples/community/fresco_v2v.py
+++ b/examples/community/fresco_v2v.py
@@ -1380,7 +1380,7 @@ def __init__(
         flow_model.eval()
         self.flow_model = flow_model
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -1413,7 +1413,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -1672,7 +1672,7 @@ def prepare_ip_adapter_image_embeds(
 
         return image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -1687,7 +1687,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -1699,7 +1699,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/latent_consistency_interpolate.py b/examples/community/latent_consistency_interpolate.py
index e8349ba317e9..0495bb75058d 100644
--- a/examples/community/latent_consistency_interpolate.py
+++ b/examples/community/latent_consistency_interpolate.py
@@ -277,7 +277,7 @@ def __init__(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -459,7 +459,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -525,7 +525,7 @@ def get_guidance_scale_embedding(self, w, embedding_dim=512, dtype=torch.float32
         assert emb.shape == (w.shape[0], embedding_dim)
         return emb
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/llm_grounded_diffusion.py b/examples/community/llm_grounded_diffusion.py
index 5bf6674a43e3..b62a0b05543b 100644
--- a/examples/community/llm_grounded_diffusion.py
+++ b/examples/community/llm_grounded_diffusion.py
@@ -1195,7 +1195,7 @@ def latent_lmd_guidance(
     # Below are methods copied from StableDiffusionPipeline
     # The design choice of not inheriting from StableDiffusionPipeline is discussed here: https://github.com/huggingface/diffusers/pull/5993#issuecomment-1834258517
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -1228,7 +1228,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -1426,7 +1426,7 @@ def encode_image(self, image, device, num_images_per_prompt):
         uncond_image_embeds = torch.zeros_like(image_embeds)
         return image_embeds, uncond_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -1441,7 +1441,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -1453,7 +1453,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -1534,17 +1534,17 @@ def get_guidance_scale_embedding(self, w, embedding_dim=512, dtype=torch.float32
         return emb
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.guidance_scale
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.guidance_scale
     def guidance_scale(self):
         return self._guidance_scale
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.guidance_rescale
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.guidance_rescale
     def guidance_rescale(self):
         return self._guidance_rescale
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.clip_skip
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.clip_skip
     def clip_skip(self):
         return self._clip_skip
 
@@ -1552,16 +1552,16 @@ def clip_skip(self):
     # of the Imagen paper: https://huggingface.co/papers/2205.11487 . `guidance_scale = 1`
     # corresponds to doing no classifier free guidance.
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.do_classifier_free_guidance
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.do_classifier_free_guidance
     def do_classifier_free_guidance(self):
         return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.cross_attention_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.cross_attention_kwargs
     def cross_attention_kwargs(self):
         return self._cross_attention_kwargs
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.num_timesteps
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.num_timesteps
     def num_timesteps(self):
         return self._num_timesteps
diff --git a/examples/community/lpw_stable_diffusion_xl.py b/examples/community/lpw_stable_diffusion_xl.py
index 95d3405df50d..10bb2a70e40e 100644
--- a/examples/community/lpw_stable_diffusion_xl.py
+++ b/examples/community/lpw_stable_diffusion_xl.py
@@ -503,7 +503,7 @@ def get_weighted_text_embeddings_sdxl(
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     """
     Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and
@@ -518,7 +518,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -532,7 +532,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -721,7 +721,7 @@ def enable_model_cpu_offload(self, gpu_id=0):
         # We'll offload the last model manually.
         self.final_offload_hook = hook
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -941,7 +941,7 @@ def encode_image(self, image, device, num_images_per_prompt, output_hidden_state
 
             return image_embeds, uncond_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/matryoshka.py b/examples/community/matryoshka.py
index 09b1d1b24465..448f97c6ef45 100644
--- a/examples/community/matryoshka.py
+++ b/examples/community/matryoshka.py
@@ -121,7 +121,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     """
     Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and
@@ -136,7 +136,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/examples/community/mixture_tiling_sdxl.py b/examples/community/mixture_tiling_sdxl.py
index 66c338b5b2a2..e3f024ec6b71 100644
--- a/examples/community/mixture_tiling_sdxl.py
+++ b/examples/community/mixture_tiling_sdxl.py
@@ -196,7 +196,7 @@ def _get_crops_coords_list(num_rows, num_cols, output_width):
     return result_list
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -223,7 +223,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -627,7 +627,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/mod_controlnet_tile_sr_sdxl.py b/examples/community/mod_controlnet_tile_sr_sdxl.py
index 27249ce3fb5e..af93e0c9b523 100644
--- a/examples/community/mod_controlnet_tile_sr_sdxl.py
+++ b/examples/community/mod_controlnet_tile_sr_sdxl.py
@@ -244,7 +244,7 @@ def _tile2latent_indices(
     return latent_row_init, latent_row_end, latent_col_init, latent_col_end
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -394,7 +394,7 @@ class TileWeightingMethod(Enum):
         COSINE = "Cosine"
         GAUSSIAN = "Gaussian"
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -633,7 +633,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_animatediff_controlnet.py b/examples/community/pipeline_animatediff_controlnet.py
index 903bfd4fd57b..0c4dc81c91cd 100644
--- a/examples/community/pipeline_animatediff_controlnet.py
+++ b/examples/community/pipeline_animatediff_controlnet.py
@@ -194,7 +194,7 @@ def __init__(
             vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True, do_normalize=False
         )
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
     def encode_prompt(
         self,
         prompt,
@@ -460,7 +460,7 @@ def decode_latents(self, latents):
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_animatediff_img2video.py b/examples/community/pipeline_animatediff_img2video.py
index feba19f70fc6..e6ad06835026 100644
--- a/examples/community/pipeline_animatediff_img2video.py
+++ b/examples/community/pipeline_animatediff_img2video.py
@@ -180,7 +180,7 @@ def tensor2vid(video: torch.Tensor, processor, output_type="np"):
     return outputs
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -194,7 +194,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -311,7 +311,7 @@ def __init__(
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
     def encode_prompt(
         self,
         prompt,
@@ -577,7 +577,7 @@ def decode_latents(self, latents):
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_animatediff_ipex.py b/examples/community/pipeline_animatediff_ipex.py
index 409ab9d6ad73..cf8a916922ae 100644
--- a/examples/community/pipeline_animatediff_ipex.py
+++ b/examples/community/pipeline_animatediff_ipex.py
@@ -165,7 +165,7 @@ def __init__(
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.video_processor = VideoProcessor(do_resize=False, vae_scale_factor=self.vae_scale_factor)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
     def encode_prompt(
         self,
         prompt,
@@ -438,7 +438,7 @@ def decode_latents(self, latents):
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_controlnet_xl_kolors.py b/examples/community/pipeline_controlnet_xl_kolors.py
index dc90aacdbc6b..a5da028532dd 100644
--- a/examples/community/pipeline_controlnet_xl_kolors.py
+++ b/examples/community/pipeline_controlnet_xl_kolors.py
@@ -113,7 +113,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -458,7 +458,7 @@ def encode_image(self, image, device, num_images_per_prompt, output_hidden_state
 
             return image_embeds, uncond_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_controlnet_xl_kolors_img2img.py b/examples/community/pipeline_controlnet_xl_kolors_img2img.py
index 189d0312143f..7f661cfa1f9d 100644
--- a/examples/community/pipeline_controlnet_xl_kolors_img2img.py
+++ b/examples/community/pipeline_controlnet_xl_kolors_img2img.py
@@ -133,7 +133,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -501,7 +501,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for others.
diff --git a/examples/community/pipeline_controlnet_xl_kolors_inpaint.py b/examples/community/pipeline_controlnet_xl_kolors_inpaint.py
index 3abd98482980..66e476183d2f 100644
--- a/examples/community/pipeline_controlnet_xl_kolors_inpaint.py
+++ b/examples/community/pipeline_controlnet_xl_kolors_inpaint.py
@@ -120,7 +120,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -134,7 +134,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -552,7 +552,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_demofusion_sdxl.py b/examples/community/pipeline_demofusion_sdxl.py
index 119b39cefe68..35d5997a0e7e 100644
--- a/examples/community/pipeline_demofusion_sdxl.py
+++ b/examples/community/pipeline_demofusion_sdxl.py
@@ -73,7 +73,7 @@ def gaussian_filter(latents, kernel_size=3, sigma=1.0):
     return blurred_latents
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     """
     Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and
@@ -379,7 +379,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_fabric.py b/examples/community/pipeline_fabric.py
index d29e98df5e50..dd3762b9436e 100644
--- a/examples/community/pipeline_fabric.py
+++ b/examples/community/pipeline_fabric.py
@@ -186,7 +186,7 @@ def __init__(
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
diff --git a/examples/community/pipeline_faithdiff_stable_diffusion_xl.py b/examples/community/pipeline_faithdiff_stable_diffusion_xl.py
index 3ebaed1fd570..70eeb351811f 100644
--- a/examples/community/pipeline_faithdiff_stable_diffusion_xl.py
+++ b/examples/community/pipeline_faithdiff_stable_diffusion_xl.py
@@ -1078,7 +1078,7 @@ def forward(self, x):
         return out
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     """
     Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and
@@ -1101,7 +1101,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -1125,7 +1125,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -1505,7 +1505,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_flux_differential_img2img.py b/examples/community/pipeline_flux_differential_img2img.py
index 3677e73136f7..569a9883e5b5 100644
--- a/examples/community/pipeline_flux_differential_img2img.py
+++ b/examples/community/pipeline_flux_differential_img2img.py
@@ -81,7 +81,7 @@
         """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -95,7 +95,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -109,7 +109,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -502,7 +502,7 @@ def check_inputs(
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._prepare_latent_image_ids
     def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
         latent_image_ids = torch.zeros(height // 2, width // 2, 3)
         latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height // 2)[:, None]
@@ -517,7 +517,7 @@ def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
         return latent_image_ids.to(device=device, dtype=dtype)
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._pack_latents
     def _pack_latents(latents, batch_size, num_channels_latents, height, width):
         latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
         latents = latents.permute(0, 2, 4, 1, 3, 5)
diff --git a/examples/community/pipeline_flux_kontext_multiple_images.py b/examples/community/pipeline_flux_kontext_multiple_images.py
index 9e6ae427dbfa..7f1d40db52c6 100644
--- a/examples/community/pipeline_flux_kontext_multiple_images.py
+++ b/examples/community/pipeline_flux_kontext_multiple_images.py
@@ -126,7 +126,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -186,7 +186,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -567,7 +567,7 @@ def check_inputs(
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._prepare_latent_image_ids
     def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
         latent_image_ids = torch.zeros(height, width, 3)
         latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
@@ -582,7 +582,7 @@ def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
         return latent_image_ids.to(device=device, dtype=dtype)
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._pack_latents
     def _pack_latents(latents, batch_size, num_channels_latents, height, width):
         latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
         latents = latents.permute(0, 2, 4, 1, 3, 5)
diff --git a/examples/community/pipeline_flux_rf_inversion.py b/examples/community/pipeline_flux_rf_inversion.py
index 2cd6eb088cd8..6d78147f65da 100644
--- a/examples/community/pipeline_flux_rf_inversion.py
+++ b/examples/community/pipeline_flux_rf_inversion.py
@@ -89,7 +89,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -103,7 +103,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/examples/community/pipeline_flux_semantic_guidance.py b/examples/community/pipeline_flux_semantic_guidance.py
index 74cd5c6981b0..87f66cb83712 100644
--- a/examples/community/pipeline_flux_semantic_guidance.py
+++ b/examples/community/pipeline_flux_semantic_guidance.py
@@ -86,7 +86,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -100,7 +100,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -640,7 +640,7 @@ def check_inputs(
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._prepare_latent_image_ids
     def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
         latent_image_ids = torch.zeros(height, width, 3)
         latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
@@ -655,7 +655,7 @@ def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
         return latent_image_ids.to(device=device, dtype=dtype)
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._pack_latents
     def _pack_latents(latents, batch_size, num_channels_latents, height, width):
         latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
         latents = latents.permute(0, 2, 4, 1, 3, 5)
diff --git a/examples/community/pipeline_flux_with_cfg.py b/examples/community/pipeline_flux_with_cfg.py
index 5bc13f7e5e11..5afd6df2deec 100644
--- a/examples/community/pipeline_flux_with_cfg.py
+++ b/examples/community/pipeline_flux_with_cfg.py
@@ -65,7 +65,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -79,7 +79,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/examples/community/pipeline_hunyuandit_differential_img2img.py b/examples/community/pipeline_hunyuandit_differential_img2img.py
index fb7a4cb5e472..15bd623a099c 100644
--- a/examples/community/pipeline_hunyuandit_differential_img2img.py
+++ b/examples/community/pipeline_hunyuandit_differential_img2img.py
@@ -146,7 +146,7 @@ def get_resize_crop_region_for_grid(src, tgt_size):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     """
     Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and
@@ -161,7 +161,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor,
     generator: Optional[torch.Generator] = None,
@@ -177,7 +177,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -512,7 +512,7 @@ def encode_prompt(
             negative_prompt_attention_mask,
         )
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -527,7 +527,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_kolors_differential_img2img.py b/examples/community/pipeline_kolors_differential_img2img.py
index d299c839815e..753595b543e2 100644
--- a/examples/community/pipeline_kolors_differential_img2img.py
+++ b/examples/community/pipeline_kolors_differential_img2img.py
@@ -66,7 +66,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -80,7 +80,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -458,7 +458,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -709,7 +709,7 @@ def _get_add_time_ids(
         add_time_ids = torch.tensor([add_time_ids], dtype=dtype)
         return add_time_ids
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.upcast_vae
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.upcast_vae
     def upcast_vae(self):
         dtype = self.vae.dtype
         self.vae.to(dtype=torch.float32)
diff --git a/examples/community/pipeline_kolors_inpainting.py b/examples/community/pipeline_kolors_inpainting.py
index 26517819ebd0..702f569b7719 100644
--- a/examples/community/pipeline_kolors_inpainting.py
+++ b/examples/community/pipeline_kolors_inpainting.py
@@ -94,7 +94,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     """
     Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and
@@ -243,7 +243,7 @@ def prepare_mask_and_masked_image(image, mask, height, width, return_image: bool
     return mask, masked_image
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -257,7 +257,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -669,7 +669,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_prompt2prompt.py b/examples/community/pipeline_prompt2prompt.py
index eb19667970b0..0d86f9ec7c03 100644
--- a/examples/community/pipeline_prompt2prompt.py
+++ b/examples/community/pipeline_prompt2prompt.py
@@ -57,7 +57,7 @@
 logger = logging.get_logger(__name__)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     """
     Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and
@@ -213,7 +213,7 @@ def __init__(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -246,7 +246,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -430,7 +430,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -445,7 +445,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_sdxl_style_aligned.py b/examples/community/pipeline_sdxl_style_aligned.py
index 51547599f531..076d6b5338d1 100644
--- a/examples/community/pipeline_sdxl_style_aligned.py
+++ b/examples/community/pipeline_sdxl_style_aligned.py
@@ -311,7 +311,7 @@ def __call__(
         return hidden_states
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     """
     Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and
@@ -326,7 +326,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -371,7 +371,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -769,7 +769,7 @@ def encode_image(self, image, device, num_images_per_prompt, output_hidden_state
 
             return image_embeds, uncond_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_stable_diffusion_3_differential_img2img.py b/examples/community/pipeline_stable_diffusion_3_differential_img2img.py
index 1803cf60cc4b..7fa6e45415b1 100644
--- a/examples/community/pipeline_stable_diffusion_3_differential_img2img.py
+++ b/examples/community/pipeline_stable_diffusion_3_differential_img2img.py
@@ -66,7 +66,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -80,7 +80,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/examples/community/pipeline_stable_diffusion_3_instruct_pix2pix.py b/examples/community/pipeline_stable_diffusion_3_instruct_pix2pix.py
index d9cee800e8ad..67728d76dd73 100644
--- a/examples/community/pipeline_stable_diffusion_3_instruct_pix2pix.py
+++ b/examples/community/pipeline_stable_diffusion_3_instruct_pix2pix.py
@@ -86,7 +86,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -100,7 +100,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -114,7 +114,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/examples/community/pipeline_stable_diffusion_upscale_ldm3d.py b/examples/community/pipeline_stable_diffusion_upscale_ldm3d.py
index e358f66b4acc..2a705be6c333 100644
--- a/examples/community/pipeline_stable_diffusion_upscale_ldm3d.py
+++ b/examples/community/pipeline_stable_diffusion_upscale_ldm3d.py
@@ -386,7 +386,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_stable_diffusion_xl_attentive_eraser.py b/examples/community/pipeline_stable_diffusion_xl_attentive_eraser.py
index 65df4c03eb2c..1dd52d0446b1 100644
--- a/examples/community/pipeline_stable_diffusion_xl_attentive_eraser.py
+++ b/examples/community/pipeline_stable_diffusion_xl_attentive_eraser.py
@@ -297,7 +297,7 @@ def forward(self, q, k, v, sim, attn, is_cross, place_in_unet, num_heads, **kwar
         return out
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     """
     Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and
@@ -439,7 +439,7 @@ def prepare_mask_and_masked_image(image, mask, height, width, return_image: bool
     return mask, masked_image
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -453,7 +453,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -696,7 +696,7 @@ def prepare_ip_adapter_image_embeds(
 
         return image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -931,7 +931,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_stable_diffusion_xl_controlnet_adapter.py b/examples/community/pipeline_stable_diffusion_xl_controlnet_adapter.py
index 6dade126f282..f4f905c9920f 100644
--- a/examples/community/pipeline_stable_diffusion_xl_controlnet_adapter.py
+++ b/examples/community/pipeline_stable_diffusion_xl_controlnet_adapter.py
@@ -137,7 +137,7 @@ def _preprocess_adapter_image(image, height, width):
     return image
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     """
     Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and
@@ -237,7 +237,7 @@ def __init__(
             else 128
         )
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -475,7 +475,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_stable_diffusion_xl_controlnet_adapter_inpaint.py b/examples/community/pipeline_stable_diffusion_xl_controlnet_adapter_inpaint.py
index 9ec6a90b072d..823e317f51a0 100644
--- a/examples/community/pipeline_stable_diffusion_xl_controlnet_adapter_inpaint.py
+++ b/examples/community/pipeline_stable_diffusion_xl_controlnet_adapter_inpaint.py
@@ -283,7 +283,7 @@ def prepare_mask_and_masked_image(image, mask, height, width, return_image: bool
     return mask, masked_image
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     """
     Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and
@@ -384,7 +384,7 @@ def __init__(
             else 128
         )
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -622,7 +622,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_stable_diffusion_xl_differential_img2img.py b/examples/community/pipeline_stable_diffusion_xl_differential_img2img.py
index 89388e10cb19..47d39ed47878 100644
--- a/examples/community/pipeline_stable_diffusion_xl_differential_img2img.py
+++ b/examples/community/pipeline_stable_diffusion_xl_differential_img2img.py
@@ -88,7 +88,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     """
     Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and
@@ -103,7 +103,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -117,7 +117,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -268,7 +268,7 @@ def __init__(
         else:
             self.watermark = None
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -506,7 +506,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_stable_diffusion_xl_ipex.py b/examples/community/pipeline_stable_diffusion_xl_ipex.py
index aa2b24f3965a..c61107cce143 100644
--- a/examples/community/pipeline_stable_diffusion_xl_ipex.py
+++ b/examples/community/pipeline_stable_diffusion_xl_ipex.py
@@ -98,7 +98,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     """
     Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and
@@ -113,7 +113,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -520,7 +520,7 @@ def encode_image(self, image, device, num_images_per_prompt):
         uncond_image_embeds = torch.zeros_like(image_embeds)
         return image_embeds, uncond_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_stg_cogvideox.py b/examples/community/pipeline_stg_cogvideox.py
index bdb6aecc30c3..c8d1a476c28a 100644
--- a/examples/community/pipeline_stg_cogvideox.py
+++ b/examples/community/pipeline_stg_cogvideox.py
@@ -138,7 +138,7 @@ def get_resize_crop_region_for_grid(src, tgt_width, tgt_height):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -409,7 +409,7 @@ def decode_latents(self, latents: torch.Tensor) -> torch.Tensor:
         frames = self.vae.decode(latents).sample
         return frames
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/pipeline_stg_hunyuan_video.py b/examples/community/pipeline_stg_hunyuan_video.py
index 028d54d047e4..eb6fc6075b2c 100644
--- a/examples/community/pipeline_stg_hunyuan_video.py
+++ b/examples/community/pipeline_stg_hunyuan_video.py
@@ -142,7 +142,7 @@ def forward_without_stg(
     return hidden_states, encoder_hidden_states
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/examples/community/pipeline_stg_ltx.py b/examples/community/pipeline_stg_ltx.py
index 70069a33f5d9..bfcd9df25312 100644
--- a/examples/community/pipeline_stg_ltx.py
+++ b/examples/community/pipeline_stg_ltx.py
@@ -119,7 +119,7 @@ def forward_with_stg(
     return hidden_states
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -133,7 +133,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/examples/community/pipeline_stg_ltx_image2video.py b/examples/community/pipeline_stg_ltx_image2video.py
index c32805e1419f..c02b102d2dd2 100644
--- a/examples/community/pipeline_stg_ltx_image2video.py
+++ b/examples/community/pipeline_stg_ltx_image2video.py
@@ -124,7 +124,7 @@ def forward_with_stg(
     return hidden_states
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -138,7 +138,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -198,7 +198,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/examples/community/pipeline_stg_mochi.py b/examples/community/pipeline_stg_mochi.py
index ad9317f6bc9d..90418941ac46 100644
--- a/examples/community/pipeline_stg_mochi.py
+++ b/examples/community/pipeline_stg_mochi.py
@@ -137,7 +137,7 @@ def linear_quadratic_schedule(num_steps, threshold_noise, linear_steps=None):
     return sigma_schedule
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/examples/community/pipline_flux_fill_controlnet_Inpaint.py b/examples/community/pipline_flux_fill_controlnet_Inpaint.py
index cc642a767f87..278a5d89beca 100644
--- a/examples/community/pipline_flux_fill_controlnet_Inpaint.py
+++ b/examples/community/pipline_flux_fill_controlnet_Inpaint.py
@@ -83,7 +83,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -97,7 +97,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -124,7 +124,7 @@ def retrieve_latents_fill(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -534,7 +534,7 @@ def check_inputs(
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._prepare_latent_image_ids
     def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
         latent_image_ids = torch.zeros(height, width, 3)
         latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
@@ -549,7 +549,7 @@ def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
         return latent_image_ids.to(device=device, dtype=dtype)
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._pack_latents
     def _pack_latents(latents, batch_size, num_channels_latents, height, width):
         latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
         latents = latents.permute(0, 2, 4, 1, 3, 5)
diff --git a/examples/community/run_onnx_controlnet.py b/examples/community/run_onnx_controlnet.py
index 2b56e8a1e51b..79ba033c841f 100644
--- a/examples/community/run_onnx_controlnet.py
+++ b/examples/community/run_onnx_controlnet.py
@@ -234,7 +234,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         warnings.warn(
             "The decode_latents method is deprecated and will be removed in a future version. Please"
@@ -248,7 +248,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/run_tensorrt_controlnet.py b/examples/community/run_tensorrt_controlnet.py
index b62eb4f58e8f..7b63be37c601 100644
--- a/examples/community/run_tensorrt_controlnet.py
+++ b/examples/community/run_tensorrt_controlnet.py
@@ -338,7 +338,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         warnings.warn(
             "The decode_latents method is deprecated and will be removed in a future version. Please"
@@ -352,7 +352,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/community/stable_diffusion_reference.py b/examples/community/stable_diffusion_reference.py
index d0372bbeba65..6b1dc4c69044 100644
--- a/examples/community/stable_diffusion_reference.py
+++ b/examples/community/stable_diffusion_reference.py
@@ -358,7 +358,7 @@ def check_inputs(
                 "Provide either `ip_adapter_image` or `ip_adapter_image_embeds`. Cannot leave both `ip_adapter_image` and `ip_adapter_image_embeds` defined."
             )
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt: Union[str, List[str]],
@@ -408,7 +408,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: Optional[str],
@@ -639,7 +639,7 @@ def prepare_latents(
         latents = latents * self.scheduler.init_noise_sigma
         return latents
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(
         self, generator: Union[torch.Generator, List[torch.Generator]], eta: float
     ) -> Dict[str, Any]:
@@ -789,7 +789,7 @@ def prepare_ref_latents(
         ref_image_latents = ref_image_latents.to(device=device, dtype=dtype)
         return ref_image_latents
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(
         self, image: Union[torch.Tensor, PIL.Image.Image], device: torch.device, dtype: torch.dtype
     ) -> Tuple[Union[torch.Tensor, PIL.Image.Image], Optional[bool]]:
diff --git a/examples/community/stable_diffusion_repaint.py b/examples/community/stable_diffusion_repaint.py
index b974e3c7ae74..cf386c0334f5 100644
--- a/examples/community/stable_diffusion_repaint.py
+++ b/examples/community/stable_diffusion_repaint.py
@@ -281,7 +281,7 @@ def __init__(
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -427,7 +427,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is not None:
             safety_checker_input = self.feature_extractor(self.numpy_to_pil(image), return_tensors="pt").to(device)
@@ -438,7 +438,7 @@ def run_safety_checker(self, image, device, dtype):
             has_nsfw_concept = None
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -456,7 +456,7 @@ def prepare_extra_step_kwargs(self, generator, eta):
             extra_step_kwargs["generator"] = generator
         return extra_step_kwargs
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         latents = 1 / self.vae.config.scaling_factor * latents
         image = self.vae.decode(latents).sample
diff --git a/examples/community/stable_diffusion_tensorrt_img2img.py b/examples/community/stable_diffusion_tensorrt_img2img.py
index 5b7733fe5784..4e281febf68f 100755
--- a/examples/community/stable_diffusion_tensorrt_img2img.py
+++ b/examples/community/stable_diffusion_tensorrt_img2img.py
@@ -832,7 +832,7 @@ def __loadModels(self):
         if "vae_encoder" in self.stages:
             self.models["vae_encoder"] = make_VAEEncoder(self.vae, **models_args)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(
         self, image: Union[torch.Tensor, PIL.Image.Image], device: torch.device, dtype: torch.dtype
     ) -> Tuple[Union[torch.Tensor, PIL.Image.Image], Optional[bool]]:
diff --git a/examples/community/stable_diffusion_tensorrt_inpaint.py b/examples/community/stable_diffusion_tensorrt_inpaint.py
index fc81e4c28917..4123f674731d 100755
--- a/examples/community/stable_diffusion_tensorrt_inpaint.py
+++ b/examples/community/stable_diffusion_tensorrt_inpaint.py
@@ -915,7 +915,7 @@ def prepare_latents(
 
         return outputs
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(
         self, image: Union[torch.Tensor, PIL.Image.Image], device: torch.device, dtype: torch.dtype
     ) -> Tuple[Union[torch.Tensor, PIL.Image.Image], Optional[bool]]:
diff --git a/examples/community/stable_diffusion_tensorrt_txt2img.py b/examples/community/stable_diffusion_tensorrt_txt2img.py
index e1d09edf9342..d79e66e9524b 100755
--- a/examples/community/stable_diffusion_tensorrt_txt2img.py
+++ b/examples/community/stable_diffusion_tensorrt_txt2img.py
@@ -788,7 +788,7 @@ def prepare_latents(
         latents = latents * self.scheduler.init_noise_sigma
         return latents
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(
         self, image: Union[torch.Tensor, PIL.Image.Image], device: torch.device, dtype: torch.dtype
     ) -> Tuple[Union[torch.Tensor, PIL.Image.Image], Optional[bool]]:
diff --git a/examples/community/stable_diffusion_xl_controlnet_reference.py b/examples/community/stable_diffusion_xl_controlnet_reference.py
index 421e67f5bba6..dc365c578793 100644
--- a/examples/community/stable_diffusion_xl_controlnet_reference.py
+++ b/examples/community/stable_diffusion_xl_controlnet_reference.py
@@ -87,7 +87,7 @@ def torch_dfs(model: torch.nn.Module):
     return result
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/examples/community/stable_diffusion_xl_reference.py b/examples/community/stable_diffusion_xl_reference.py
index eb055574966d..6802c4af207b 100644
--- a/examples/community/stable_diffusion_xl_reference.py
+++ b/examples/community/stable_diffusion_xl_reference.py
@@ -61,7 +61,7 @@ def torch_dfs(model: torch.nn.Module):
     return result
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     """
     Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and
@@ -76,7 +76,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/examples/research_projects/anytext/anytext.py b/examples/research_projects/anytext/anytext.py
index 7ae6ae57c22a..40da856e8184 100644
--- a/examples/research_projects/anytext/anytext.py
+++ b/examples/research_projects/anytext/anytext.py
@@ -1016,7 +1016,7 @@ def insert_spaces(self, string, nSpace):
         return new_string[:-nSpace]
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -1124,7 +1124,7 @@ def insert_spaces(self, string, nSpace):
         return new_string[:-nSpace]
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -1323,7 +1323,7 @@ def is_chinese(self, text):
                 return True
         return False
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -1356,7 +1356,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -1610,7 +1610,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -1625,7 +1625,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -1637,7 +1637,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/research_projects/pixart/pipeline_pixart_alpha_controlnet.py b/examples/research_projects/pixart/pipeline_pixart_alpha_controlnet.py
index 89228983d4d8..93c9a505a057 100644
--- a/examples/research_projects/pixart/pipeline_pixart_alpha_controlnet.py
+++ b/examples/research_projects/pixart/pipeline_pixart_alpha_controlnet.py
@@ -185,7 +185,7 @@ def get_closest_hw(width, height, image_size):
     return width, height
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -457,7 +457,7 @@ def encode_prompt(
 
         return prompt_embeds, prompt_attention_mask, negative_prompt_embeds, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/examples/research_projects/promptdiffusion/pipeline_prompt_diffusion.py b/examples/research_projects/promptdiffusion/pipeline_prompt_diffusion.py
index 233df1276563..5b7ac9ec8a30 100644
--- a/examples/research_projects/promptdiffusion/pipeline_prompt_diffusion.py
+++ b/examples/research_projects/promptdiffusion/pipeline_prompt_diffusion.py
@@ -97,7 +97,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -285,7 +285,7 @@ def disable_vae_tiling(self):
         )
         self.vae.disable_tiling()
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -318,7 +318,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -525,7 +525,7 @@ def encode_image(self, image, device, num_images_per_prompt, output_hidden_state
 
             return image_embeds, uncond_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -540,7 +540,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -552,7 +552,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/modular_pipelines/flux/before_denoise.py b/src/diffusers/modular_pipelines/flux/before_denoise.py
index daffec986535..c719cd28e005 100644
--- a/src/diffusers/modular_pipelines/flux/before_denoise.py
+++ b/src/diffusers/modular_pipelines/flux/before_denoise.py
@@ -19,6 +19,7 @@
 import torch
 
 from ...pipelines import FluxPipeline
+from ...pipelines.flux.pipeline_flux_utils import calculate_shift
 from ...schedulers import FlowMatchEulerDiscreteScheduler
 from ...utils import logging
 from ...utils.torch_utils import randn_tensor
@@ -30,7 +31,7 @@
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -90,21 +91,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/modular_pipelines/flux/encoders.py b/src/diffusers/modular_pipelines/flux/encoders.py
index f0314d4771b0..c9f72e2a6993 100644
--- a/src/diffusers/modular_pipelines/flux/encoders.py
+++ b/src/diffusers/modular_pipelines/flux/encoders.py
@@ -53,7 +53,7 @@ def prompt_clean(text):
     return text
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/modular_pipelines/qwenimage/before_denoise.py b/src/diffusers/modular_pipelines/qwenimage/before_denoise.py
index 0e470332c6f4..eb96cda7dba0 100644
--- a/src/diffusers/modular_pipelines/qwenimage/before_denoise.py
+++ b/src/diffusers/modular_pipelines/qwenimage/before_denoise.py
@@ -26,7 +26,7 @@
 from .modular_pipeline import QwenImageModularPipeline, QwenImagePachifier
 
 
-# Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.calculate_shift
+# Copied from diffusers.pipelines.qwenimage.pipeline_qwen_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -40,7 +40,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/modular_pipelines/qwenimage/encoders.py b/src/diffusers/modular_pipelines/qwenimage/encoders.py
index 3b56981e5290..be2dbf938641 100644
--- a/src/diffusers/modular_pipelines/qwenimage/encoders.py
+++ b/src/diffusers/modular_pipelines/qwenimage/encoders.py
@@ -183,7 +183,7 @@ def get_qwen_prompt_embeds_edit_plus(
     return prompt_embeds, encoder_attention_mask
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/modular_pipelines/stable_diffusion_xl/before_denoise.py b/src/diffusers/modular_pipelines/stable_diffusion_xl/before_denoise.py
index 70cbf0c1c78d..bd7189105a36 100644
--- a/src/diffusers/modular_pipelines/stable_diffusion_xl/before_denoise.py
+++ b/src/diffusers/modular_pipelines/stable_diffusion_xl/before_denoise.py
@@ -43,7 +43,7 @@
 # configuration of guider is.
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -103,7 +103,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/modular_pipelines/stable_diffusion_xl/decoders.py b/src/diffusers/modular_pipelines/stable_diffusion_xl/decoders.py
index feb78e1ef11b..1a0b36a91c13 100644
--- a/src/diffusers/modular_pipelines/stable_diffusion_xl/decoders.py
+++ b/src/diffusers/modular_pipelines/stable_diffusion_xl/decoders.py
@@ -81,10 +81,7 @@ def upcast_vae(components):
         components.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             components.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/modular_pipelines/stable_diffusion_xl/encoders.py b/src/diffusers/modular_pipelines/stable_diffusion_xl/encoders.py
index 90b254b6f5d4..37cc77646147 100644
--- a/src/diffusers/modular_pipelines/stable_diffusion_xl/encoders.py
+++ b/src/diffusers/modular_pipelines/stable_diffusion_xl/encoders.py
@@ -43,7 +43,7 @@
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/modular_pipelines/wan/before_denoise.py b/src/diffusers/modular_pipelines/wan/before_denoise.py
index e2f8d3e7d88b..6de031935c4e 100644
--- a/src/diffusers/modular_pipelines/wan/before_denoise.py
+++ b/src/diffusers/modular_pipelines/wan/before_denoise.py
@@ -126,7 +126,7 @@ def calculate_dimension_from_latents(
     return num_frames, height, width
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/modular_pipelines/wan/encoders.py b/src/diffusers/modular_pipelines/wan/encoders.py
index dc49df8eab8c..c23a1a02ac7b 100644
--- a/src/diffusers/modular_pipelines/wan/encoders.py
+++ b/src/diffusers/modular_pipelines/wan/encoders.py
@@ -102,7 +102,7 @@ def encode_image(
     return image_embeds.hidden_states[-2]
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/pipelines/allegro/pipeline_allegro.py b/src/diffusers/pipelines/allegro/pipeline_allegro.py
index 3be0129088fb..90706007c98a 100644
--- a/src/diffusers/pipelines/allegro/pipeline_allegro.py
+++ b/src/diffusers/pipelines/allegro/pipeline_allegro.py
@@ -81,7 +81,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -347,7 +347,7 @@ def encode_prompt(
 
         return prompt_embeds, prompt_attention_mask, negative_prompt_embeds, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/animatediff/pipeline_animatediff.py b/src/diffusers/pipelines/animatediff/pipeline_animatediff.py
index 091b6db713ba..5e2e7ca61950 100644
--- a/src/diffusers/pipelines/animatediff/pipeline_animatediff.py
+++ b/src/diffusers/pipelines/animatediff/pipeline_animatediff.py
@@ -152,7 +152,7 @@ def __init__(
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.video_processor = VideoProcessor(do_resize=False, vae_scale_factor=self.vae_scale_factor)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
     def encode_prompt(
         self,
         prompt,
@@ -424,7 +424,7 @@ def decode_latents(self, latents, decode_chunk_size: int = 16):
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/animatediff/pipeline_animatediff_controlnet.py b/src/diffusers/pipelines/animatediff/pipeline_animatediff_controlnet.py
index 70180ccf0650..42344dffa7ec 100644
--- a/src/diffusers/pipelines/animatediff/pipeline_animatediff_controlnet.py
+++ b/src/diffusers/pipelines/animatediff/pipeline_animatediff_controlnet.py
@@ -195,7 +195,7 @@ def __init__(
             vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True, do_normalize=False
         )
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
     def encode_prompt(
         self,
         prompt,
@@ -468,7 +468,7 @@ def decode_latents(self, latents, decode_chunk_size: int = 16):
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/animatediff/pipeline_animatediff_sdxl.py b/src/diffusers/pipelines/animatediff/pipeline_animatediff_sdxl.py
index 56d319027595..4a76428ce4f4 100644
--- a/src/diffusers/pipelines/animatediff/pipeline_animatediff_sdxl.py
+++ b/src/diffusers/pipelines/animatediff/pipeline_animatediff_sdxl.py
@@ -120,7 +120,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -147,7 +147,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -325,7 +325,7 @@ def __init__(
             else 128
         )
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt with num_images_per_prompt->num_videos_per_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt with num_images_per_prompt->num_videos_per_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -648,7 +648,7 @@ def decode_latents(self, latents):
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/animatediff/pipeline_animatediff_sparsectrl.py b/src/diffusers/pipelines/animatediff/pipeline_animatediff_sparsectrl.py
index 46d650efe8b6..778b724cd400 100644
--- a/src/diffusers/pipelines/animatediff/pipeline_animatediff_sparsectrl.py
+++ b/src/diffusers/pipelines/animatediff/pipeline_animatediff_sparsectrl.py
@@ -115,7 +115,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -204,7 +204,7 @@ def __init__(
             vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True, do_normalize=False
         )
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
     def encode_prompt(
         self,
         prompt,
@@ -471,7 +471,7 @@ def decode_latents(self, latents):
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/animatediff/pipeline_animatediff_video2video.py b/src/diffusers/pipelines/animatediff/pipeline_animatediff_video2video.py
index 6f3a609aba4a..64c15405d0e7 100644
--- a/src/diffusers/pipelines/animatediff/pipeline_animatediff_video2video.py
+++ b/src/diffusers/pipelines/animatediff/pipeline_animatediff_video2video.py
@@ -104,7 +104,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -118,7 +118,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -535,7 +535,7 @@ def decode_latents(self, latents, decode_chunk_size: int = 16):
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/animatediff/pipeline_animatediff_video2video_controlnet.py b/src/diffusers/pipelines/animatediff/pipeline_animatediff_video2video_controlnet.py
index b00f344598ad..0f6aaa4f2c45 100644
--- a/src/diffusers/pipelines/animatediff/pipeline_animatediff_video2video_controlnet.py
+++ b/src/diffusers/pipelines/animatediff/pipeline_animatediff_video2video_controlnet.py
@@ -122,7 +122,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -136,7 +136,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -567,7 +567,7 @@ def decode_latents(self, latents, decode_chunk_size: int = 16):
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/audioldm/pipeline_audioldm.py b/src/diffusers/pipelines/audioldm/pipeline_audioldm.py
index 6a70f00c76c7..45e16bd90dfe 100644
--- a/src/diffusers/pipelines/audioldm/pipeline_audioldm.py
+++ b/src/diffusers/pipelines/audioldm/pipeline_audioldm.py
@@ -258,7 +258,7 @@ def mel_spectrogram_to_waveform(self, mel_spectrogram):
         waveform = waveform.cpu().float()
         return waveform
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/audioldm2/pipeline_audioldm2.py b/src/diffusers/pipelines/audioldm2/pipeline_audioldm2.py
index 452fc3c01b27..8e82e1c5b120 100644
--- a/src/diffusers/pipelines/audioldm2/pipeline_audioldm2.py
+++ b/src/diffusers/pipelines/audioldm2/pipeline_audioldm2.py
@@ -719,7 +719,7 @@ def score_waveforms(self, text, audio, num_waveforms_per_prompt, device, dtype):
         audio = torch.index_select(audio, 0, indices.reshape(-1).cpu())
         return audio
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/aura_flow/pipeline_aura_flow.py b/src/diffusers/pipelines/aura_flow/pipeline_aura_flow.py
index 6251ca443533..b31ff4d37813 100644
--- a/src/diffusers/pipelines/aura_flow/pipeline_aura_flow.py
+++ b/src/diffusers/pipelines/aura_flow/pipeline_aura_flow.py
@@ -60,7 +60,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -404,17 +404,13 @@ def prepare_latents(
 
         return latents
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.upcast_vae
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.upcast_vae
     def upcast_vae(self):
         dtype = self.vae.dtype
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-                FusedAttnProcessor2_0,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor, FusedAttnProcessor2_0),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/pipelines/bria_fibo/pipeline_bria_fibo.py b/src/diffusers/pipelines/bria_fibo/pipeline_bria_fibo.py
index 8fd29756b290..44e1ad56edca 100644
--- a/src/diffusers/pipelines/bria_fibo/pipeline_bria_fibo.py
+++ b/src/diffusers/pipelines/bria_fibo/pipeline_bria_fibo.py
@@ -367,7 +367,7 @@ def _unpack_latents(latents, height, width, vae_scale_factor):
         return latents
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._prepare_latent_image_ids
     def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
         latent_image_ids = torch.zeros(height, width, 3)
         latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
@@ -400,7 +400,7 @@ def _pack_latents_no_patch(latents, batch_size, num_channels_latents, height, wi
         return latents
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._pack_latents
     def _pack_latents(latents, batch_size, num_channels_latents, height, width):
         latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
         latents = latents.permute(0, 2, 4, 1, 3, 5)
diff --git a/src/diffusers/pipelines/chroma/pipeline_chroma.py b/src/diffusers/pipelines/chroma/pipeline_chroma.py
index ed6c2c2105b6..8c73d9394e4d 100644
--- a/src/diffusers/pipelines/chroma/pipeline_chroma.py
+++ b/src/diffusers/pipelines/chroma/pipeline_chroma.py
@@ -74,7 +74,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -88,7 +88,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/chroma/pipeline_chroma_img2img.py b/src/diffusers/pipelines/chroma/pipeline_chroma_img2img.py
index 470c746e4146..6e4e1bc72052 100644
--- a/src/diffusers/pipelines/chroma/pipeline_chroma_img2img.py
+++ b/src/diffusers/pipelines/chroma/pipeline_chroma_img2img.py
@@ -72,7 +72,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -86,7 +86,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -100,7 +100,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/chronoedit/pipeline_chronoedit.py b/src/diffusers/pipelines/chronoedit/pipeline_chronoedit.py
index 79f6580fbed6..f73421e38927 100644
--- a/src/diffusers/pipelines/chronoedit/pipeline_chronoedit.py
+++ b/src/diffusers/pipelines/chronoedit/pipeline_chronoedit.py
@@ -111,7 +111,7 @@ def prompt_clean(text):
     return text
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/pipelines/cogvideo/pipeline_cogvideox.py b/src/diffusers/pipelines/cogvideo/pipeline_cogvideox.py
index 4ac33b24bbe1..98101119ccfa 100644
--- a/src/diffusers/pipelines/cogvideo/pipeline_cogvideox.py
+++ b/src/diffusers/pipelines/cogvideo/pipeline_cogvideox.py
@@ -84,7 +84,7 @@ def get_resize_crop_region_for_grid(src, tgt_width, tgt_height):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -355,7 +355,7 @@ def decode_latents(self, latents: torch.Tensor) -> torch.Tensor:
         frames = self.vae.decode(latents).sample
         return frames
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/cogvideo/pipeline_cogvideox_fun_control.py b/src/diffusers/pipelines/cogvideo/pipeline_cogvideox_fun_control.py
index c1335839f848..1a5c97604960 100644
--- a/src/diffusers/pipelines/cogvideo/pipeline_cogvideox_fun_control.py
+++ b/src/diffusers/pipelines/cogvideo/pipeline_cogvideox_fun_control.py
@@ -91,7 +91,7 @@ def get_resize_crop_region_for_grid(src, tgt_width, tgt_height):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -394,7 +394,7 @@ def decode_latents(self, latents: torch.Tensor) -> torch.Tensor:
         frames = self.vae.decode(latents).sample
         return frames
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/cogvideo/pipeline_cogvideox_image2video.py b/src/diffusers/pipelines/cogvideo/pipeline_cogvideox_image2video.py
index c523c9adec98..707506c22365 100644
--- a/src/diffusers/pipelines/cogvideo/pipeline_cogvideox_image2video.py
+++ b/src/diffusers/pipelines/cogvideo/pipeline_cogvideox_image2video.py
@@ -83,7 +83,7 @@ def get_resize_crop_region_for_grid(src, tgt_width, tgt_height):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -143,7 +143,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -434,7 +434,7 @@ def get_timesteps(self, num_inference_steps, timesteps, strength, device):
 
         return timesteps, num_inference_steps - t_start
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/cogvideo/pipeline_cogvideox_video2video.py b/src/diffusers/pipelines/cogvideo/pipeline_cogvideox_video2video.py
index 897dc6d1b70a..cbd2e22c0486 100644
--- a/src/diffusers/pipelines/cogvideo/pipeline_cogvideox_video2video.py
+++ b/src/diffusers/pipelines/cogvideo/pipeline_cogvideox_video2video.py
@@ -92,7 +92,7 @@ def get_resize_crop_region_for_grid(src, tgt_width, tgt_height):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -152,7 +152,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -414,7 +414,7 @@ def get_timesteps(self, num_inference_steps, timesteps, strength, device):
 
         return timesteps, num_inference_steps - t_start
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/cogview3/pipeline_cogview3plus.py b/src/diffusers/pipelines/cogview3/pipeline_cogview3plus.py
index 304a5c5ad00b..b558a909c3f2 100644
--- a/src/diffusers/pipelines/cogview3/pipeline_cogview3plus.py
+++ b/src/diffusers/pipelines/cogview3/pipeline_cogview3plus.py
@@ -55,7 +55,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -315,7 +315,7 @@ def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype
         latents = latents * self.scheduler.init_noise_sigma
         return latents
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/consisid/pipeline_consisid.py b/src/diffusers/pipelines/consisid/pipeline_consisid.py
index 3e6c149d7f80..1214ac81e4e4 100644
--- a/src/diffusers/pipelines/consisid/pipeline_consisid.py
+++ b/src/diffusers/pipelines/consisid/pipeline_consisid.py
@@ -173,7 +173,7 @@ def get_resize_crop_region_for_grid(src, tgt_width, tgt_height):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -233,7 +233,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -536,7 +536,7 @@ def get_timesteps(self, num_inference_steps, timesteps, strength, device):
 
         return timesteps, num_inference_steps - t_start
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/controlnet/pipeline_controlnet.py b/src/diffusers/pipelines/controlnet/pipeline_controlnet.py
index fe0e69314cca..b9a6aa60b255 100644
--- a/src/diffusers/pipelines/controlnet/pipeline_controlnet.py
+++ b/src/diffusers/pipelines/controlnet/pipeline_controlnet.py
@@ -99,7 +99,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -261,7 +261,7 @@ def __init__(
         )
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -294,7 +294,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -548,7 +548,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -563,7 +563,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -575,7 +575,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/controlnet/pipeline_controlnet_img2img.py b/src/diffusers/pipelines/controlnet/pipeline_controlnet_img2img.py
index 12cc6f630d80..9766aa8557c9 100644
--- a/src/diffusers/pipelines/controlnet/pipeline_controlnet_img2img.py
+++ b/src/diffusers/pipelines/controlnet/pipeline_controlnet_img2img.py
@@ -99,7 +99,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -239,7 +239,7 @@ def __init__(
         )
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -272,7 +272,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -526,7 +526,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -541,7 +541,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -553,7 +553,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/controlnet/pipeline_controlnet_inpaint.py b/src/diffusers/pipelines/controlnet/pipeline_controlnet_inpaint.py
index 6de8e5747b02..887ffe73dc55 100644
--- a/src/diffusers/pipelines/controlnet/pipeline_controlnet_inpaint.py
+++ b/src/diffusers/pipelines/controlnet/pipeline_controlnet_inpaint.py
@@ -111,7 +111,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -245,7 +245,7 @@ def __init__(
         )
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -278,7 +278,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -532,7 +532,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -547,7 +547,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -559,7 +559,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/controlnet/pipeline_controlnet_inpaint_sd_xl.py b/src/diffusers/pipelines/controlnet/pipeline_controlnet_inpaint_sd_xl.py
index 397ab15715c2..97b75a17e50b 100644
--- a/src/diffusers/pipelines/controlnet/pipeline_controlnet_inpaint_sd_xl.py
+++ b/src/diffusers/pipelines/controlnet/pipeline_controlnet_inpaint_sd_xl.py
@@ -73,7 +73,7 @@
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -144,7 +144,7 @@ def retrieve_latents(
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -291,7 +291,7 @@ def __init__(
         else:
             self.watermark = None
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -601,7 +601,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -1136,10 +1136,7 @@ def upcast_vae(self):
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl.py b/src/diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl.py
index 8b37d38b9099..987178a33358 100644
--- a/src/diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl.py
+++ b/src/diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl.py
@@ -121,7 +121,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -299,7 +299,7 @@ def __init__(
 
         self.register_to_config(force_zeros_for_empty_prompt=force_zeros_for_empty_prompt)
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -609,7 +609,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -933,10 +933,7 @@ def upcast_vae(self):
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl_img2img.py b/src/diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl_img2img.py
index 4d4845c5a0a3..d787580e6898 100644
--- a/src/diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl_img2img.py
+++ b/src/diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl_img2img.py
@@ -150,7 +150,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -293,7 +293,7 @@ def __init__(
         self.register_to_config(force_zeros_for_empty_prompt=force_zeros_for_empty_prompt)
         self.register_to_config(requires_aesthetics_score=requires_aesthetics_score)
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -603,7 +603,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -1048,10 +1048,7 @@ def upcast_vae(self):
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/pipelines/controlnet/pipeline_controlnet_union_inpaint_sd_xl.py b/src/diffusers/pipelines/controlnet/pipeline_controlnet_union_inpaint_sd_xl.py
index fb58b222112a..6ec940b7c612 100644
--- a/src/diffusers/pipelines/controlnet/pipeline_controlnet_union_inpaint_sd_xl.py
+++ b/src/diffusers/pipelines/controlnet/pipeline_controlnet_union_inpaint_sd_xl.py
@@ -78,7 +78,7 @@
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -134,7 +134,7 @@ def retrieve_latents(
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -280,7 +280,7 @@ def __init__(
         else:
             self.watermark = None
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -590,7 +590,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -1115,10 +1115,7 @@ def upcast_vae(self):
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/pipelines/controlnet/pipeline_controlnet_union_sd_xl.py b/src/diffusers/pipelines/controlnet/pipeline_controlnet_union_sd_xl.py
index 5961d389effb..c11d7c2ff12f 100644
--- a/src/diffusers/pipelines/controlnet/pipeline_controlnet_union_sd_xl.py
+++ b/src/diffusers/pipelines/controlnet/pipeline_controlnet_union_sd_xl.py
@@ -115,7 +115,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -289,7 +289,7 @@ def __init__(
 
         self.register_to_config(force_zeros_for_empty_prompt=force_zeros_for_empty_prompt)
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -599,7 +599,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -908,10 +908,7 @@ def upcast_vae(self):
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/pipelines/controlnet/pipeline_controlnet_union_sd_xl_img2img.py b/src/diffusers/pipelines/controlnet/pipeline_controlnet_union_sd_xl_img2img.py
index 8fedb6d8609a..09b1dae8b6ee 100644
--- a/src/diffusers/pipelines/controlnet/pipeline_controlnet_union_sd_xl_img2img.py
+++ b/src/diffusers/pipelines/controlnet/pipeline_controlnet_union_sd_xl_img2img.py
@@ -175,7 +175,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -309,7 +309,7 @@ def __init__(
         self.register_to_config(force_zeros_for_empty_prompt=force_zeros_for_empty_prompt)
         self.register_to_config(requires_aesthetics_score=requires_aesthetics_score)
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -619,7 +619,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -1036,10 +1036,7 @@ def upcast_vae(self):
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/pipelines/controlnet_hunyuandit/pipeline_hunyuandit_controlnet.py b/src/diffusers/pipelines/controlnet_hunyuandit/pipeline_hunyuandit_controlnet.py
index 2b5684de9511..f8814706d906 100644
--- a/src/diffusers/pipelines/controlnet_hunyuandit/pipeline_hunyuandit_controlnet.py
+++ b/src/diffusers/pipelines/controlnet_hunyuandit/pipeline_hunyuandit_controlnet.py
@@ -135,7 +135,7 @@ def get_resize_crop_region_for_grid(src, tgt_size):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -440,7 +440,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, prompt_attention_mask, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -455,7 +455,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/controlnet_sd3/pipeline_stable_diffusion_3_controlnet.py b/src/diffusers/pipelines/controlnet_sd3/pipeline_stable_diffusion_3_controlnet.py
index d605eac1f2b1..e31dd081521f 100644
--- a/src/diffusers/pipelines/controlnet_sd3/pipeline_stable_diffusion_3_controlnet.py
+++ b/src/diffusers/pipelines/controlnet_sd3/pipeline_stable_diffusion_3_controlnet.py
@@ -80,7 +80,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/controlnet_sd3/pipeline_stable_diffusion_3_controlnet_inpainting.py b/src/diffusers/pipelines/controlnet_sd3/pipeline_stable_diffusion_3_controlnet_inpainting.py
index 9d0158c6b654..6fca057a1e74 100644
--- a/src/diffusers/pipelines/controlnet_sd3/pipeline_stable_diffusion_3_controlnet_inpainting.py
+++ b/src/diffusers/pipelines/controlnet_sd3/pipeline_stable_diffusion_3_controlnet_inpainting.py
@@ -101,7 +101,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/controlnet_xs/pipeline_controlnet_xs.py b/src/diffusers/pipelines/controlnet_xs/pipeline_controlnet_xs.py
index 3682ddc91156..26d173af7f9b 100644
--- a/src/diffusers/pipelines/controlnet_xs/pipeline_controlnet_xs.py
+++ b/src/diffusers/pipelines/controlnet_xs/pipeline_controlnet_xs.py
@@ -195,7 +195,7 @@ def __init__(
         )
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -228,7 +228,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -411,7 +411,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -426,7 +426,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -438,7 +438,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -616,27 +616,27 @@ def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype
         return latents
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.guidance_scale
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.guidance_scale
     def guidance_scale(self):
         return self._guidance_scale
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.clip_skip
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.clip_skip
     def clip_skip(self):
         return self._clip_skip
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.do_classifier_free_guidance
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.do_classifier_free_guidance
     def do_classifier_free_guidance(self):
         return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.cross_attention_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.cross_attention_kwargs
     def cross_attention_kwargs(self):
         return self._cross_attention_kwargs
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.num_timesteps
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.num_timesteps
     def num_timesteps(self):
         return self._num_timesteps
 
diff --git a/src/diffusers/pipelines/controlnet_xs/pipeline_controlnet_xs_sd_xl.py b/src/diffusers/pipelines/controlnet_xs/pipeline_controlnet_xs_sd_xl.py
index c10931a0f44a..930a5e71400f 100644
--- a/src/diffusers/pipelines/controlnet_xs/pipeline_controlnet_xs_sd_xl.py
+++ b/src/diffusers/pipelines/controlnet_xs/pipeline_controlnet_xs_sd_xl.py
@@ -222,7 +222,7 @@ def __init__(
 
         self.register_to_config(force_zeros_for_empty_prompt=force_zeros_for_empty_prompt)
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -461,7 +461,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -689,10 +689,7 @@ def upcast_vae(self):
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
@@ -702,27 +699,27 @@ def upcast_vae(self):
             self.vae.decoder.mid_block.to(dtype)
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.guidance_scale
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.guidance_scale
     def guidance_scale(self):
         return self._guidance_scale
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.clip_skip
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.clip_skip
     def clip_skip(self):
         return self._clip_skip
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.do_classifier_free_guidance
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.do_classifier_free_guidance
     def do_classifier_free_guidance(self):
         return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.cross_attention_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.cross_attention_kwargs
     def cross_attention_kwargs(self):
         return self._cross_attention_kwargs
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.num_timesteps
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.num_timesteps
     def num_timesteps(self):
         return self._num_timesteps
 
diff --git a/src/diffusers/pipelines/cosmos/pipeline_cosmos2_text2image.py b/src/diffusers/pipelines/cosmos/pipeline_cosmos2_text2image.py
index 66490c2be159..22cafd2d54e2 100644
--- a/src/diffusers/pipelines/cosmos/pipeline_cosmos2_text2image.py
+++ b/src/diffusers/pipelines/cosmos/pipeline_cosmos2_text2image.py
@@ -72,7 +72,7 @@ def __init__(self, *args, **kwargs):
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/cosmos/pipeline_cosmos2_video2world.py b/src/diffusers/pipelines/cosmos/pipeline_cosmos2_video2world.py
index 23a74ad00f93..e329810386ae 100644
--- a/src/diffusers/pipelines/cosmos/pipeline_cosmos2_video2world.py
+++ b/src/diffusers/pipelines/cosmos/pipeline_cosmos2_video2world.py
@@ -77,7 +77,7 @@ def __init__(self, *args, **kwargs):
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -137,7 +137,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/pipelines/cosmos/pipeline_cosmos_text2world.py b/src/diffusers/pipelines/cosmos/pipeline_cosmos_text2world.py
index f0aa1ecf0e0f..4aceafe63c25 100644
--- a/src/diffusers/pipelines/cosmos/pipeline_cosmos_text2world.py
+++ b/src/diffusers/pipelines/cosmos/pipeline_cosmos_text2world.py
@@ -69,7 +69,7 @@ def __init__(self, *args, **kwargs):
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/cosmos/pipeline_cosmos_video2world.py b/src/diffusers/pipelines/cosmos/pipeline_cosmos_video2world.py
index cd5a734cc311..6ea2bc959f87 100644
--- a/src/diffusers/pipelines/cosmos/pipeline_cosmos_video2world.py
+++ b/src/diffusers/pipelines/cosmos/pipeline_cosmos_video2world.py
@@ -98,7 +98,7 @@ def __init__(self, *args, **kwargs):
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -158,7 +158,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/pipelines/deepfloyd_if/pipeline_if.py b/src/diffusers/pipelines/deepfloyd_if/pipeline_if.py
index 8fa31f8504d3..b5294c1782da 100644
--- a/src/diffusers/pipelines/deepfloyd_if/pipeline_if.py
+++ b/src/diffusers/pipelines/deepfloyd_if/pipeline_if.py
@@ -332,7 +332,7 @@ def run_safety_checker(self, image, device, dtype):
 
         return image, nsfw_detected, watermark_detected
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/deprecated/alt_diffusion/pipeline_alt_diffusion.py b/src/diffusers/pipelines/deprecated/alt_diffusion/pipeline_alt_diffusion.py
index 6f484aa3e298..ed238ab861b1 100644
--- a/src/diffusers/pipelines/deprecated/alt_diffusion/pipeline_alt_diffusion.py
+++ b/src/diffusers/pipelines/deprecated/alt_diffusion/pipeline_alt_diffusion.py
@@ -63,7 +63,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -90,7 +90,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/deprecated/alt_diffusion/pipeline_alt_diffusion_img2img.py b/src/diffusers/pipelines/deprecated/alt_diffusion/pipeline_alt_diffusion_img2img.py
index d6bf90120755..95846bc6eb3f 100644
--- a/src/diffusers/pipelines/deprecated/alt_diffusion/pipeline_alt_diffusion_img2img.py
+++ b/src/diffusers/pipelines/deprecated/alt_diffusion/pipeline_alt_diffusion_img2img.py
@@ -80,7 +80,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -118,7 +118,7 @@ def preprocess(image):
     return image
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_cycle_diffusion.py b/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_cycle_diffusion.py
index 08f8c7e26fae..7884595728d5 100644
--- a/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_cycle_diffusion.py
+++ b/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_cycle_diffusion.py
@@ -61,7 +61,7 @@ def preprocess(image):
     return image
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -251,7 +251,7 @@ def __init__(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -284,7 +284,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -518,7 +518,7 @@ def check_inputs(
                     f" {negative_prompt_embeds.shape}."
                 )
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -536,7 +536,7 @@ def prepare_extra_step_kwargs(self, generator, eta):
             extra_step_kwargs["generator"] = generator
         return extra_step_kwargs
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -551,7 +551,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
diff --git a/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_inpaint_legacy.py b/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_inpaint_legacy.py
index ba0dd66c2938..af7036f1211b 100644
--- a/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_inpaint_legacy.py
+++ b/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_inpaint_legacy.py
@@ -222,7 +222,7 @@ def __init__(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -255,7 +255,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -438,7 +438,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -453,7 +453,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -465,7 +465,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_model_editing.py b/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_model_editing.py
index b7a0be57c12b..d7cc60cd57c3 100644
--- a/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_model_editing.py
+++ b/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_model_editing.py
@@ -155,7 +155,7 @@ def append_ca(net_):
             self.projection_matrices = self.projection_matrices + [l.to_k for l in self.ca_clip_layers]
             self.og_matrices = self.og_matrices + [copy.deepcopy(l.to_k) for l in self.ca_clip_layers]
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -188,7 +188,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -371,7 +371,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -386,7 +386,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -398,7 +398,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_paradigms.py b/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_paradigms.py
index c236e73bf448..2a15e01f0789 100644
--- a/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_paradigms.py
+++ b/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_paradigms.py
@@ -152,7 +152,7 @@ def __init__(
         # attribute to wrap the unet with torch.nn.DataParallel when running multiple denoising steps on multiple GPUs
         self.wrapped_unet = self.unet
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -185,7 +185,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -368,7 +368,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -383,7 +383,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_pix2pix_zero.py b/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_pix2pix_zero.py
index 2a461ae20cc9..f998b0429a1d 100644
--- a/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_pix2pix_zero.py
+++ b/src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_pix2pix_zero.py
@@ -370,7 +370,7 @@ def __init__(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -403,7 +403,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -586,7 +586,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -601,7 +601,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -613,7 +613,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/deprecated/versatile_diffusion/pipeline_versatile_diffusion_dual_guided.py b/src/diffusers/pipelines/deprecated/versatile_diffusion/pipeline_versatile_diffusion_dual_guided.py
index 0252f4f6af7f..e0e6069dbf32 100644
--- a/src/diffusers/pipelines/deprecated/versatile_diffusion/pipeline_versatile_diffusion_dual_guided.py
+++ b/src/diffusers/pipelines/deprecated/versatile_diffusion/pipeline_versatile_diffusion_dual_guided.py
@@ -299,7 +299,7 @@ def normalize_embeddings(encoder_output):
 
         return image_embeddings
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -311,7 +311,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/deprecated/versatile_diffusion/pipeline_versatile_diffusion_image_variation.py b/src/diffusers/pipelines/deprecated/versatile_diffusion/pipeline_versatile_diffusion_image_variation.py
index 034a0226419b..fdb87773b8c1 100644
--- a/src/diffusers/pipelines/deprecated/versatile_diffusion/pipeline_versatile_diffusion_image_variation.py
+++ b/src/diffusers/pipelines/deprecated/versatile_diffusion/pipeline_versatile_diffusion_image_variation.py
@@ -158,7 +158,7 @@ def normalize_embeddings(encoder_output):
 
         return image_embeddings
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -170,7 +170,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/deprecated/versatile_diffusion/pipeline_versatile_diffusion_text_to_image.py b/src/diffusers/pipelines/deprecated/versatile_diffusion/pipeline_versatile_diffusion_text_to_image.py
index 2f54f4fc98a4..94be3d760eb9 100644
--- a/src/diffusers/pipelines/deprecated/versatile_diffusion/pipeline_versatile_diffusion_text_to_image.py
+++ b/src/diffusers/pipelines/deprecated/versatile_diffusion/pipeline_versatile_diffusion_text_to_image.py
@@ -215,7 +215,7 @@ def normalize_embeddings(encoder_output):
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -227,7 +227,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/easyanimate/pipeline_easyanimate.py b/src/diffusers/pipelines/easyanimate/pipeline_easyanimate.py
index 92239c0d32f0..d318275a240b 100755
--- a/src/diffusers/pipelines/easyanimate/pipeline_easyanimate.py
+++ b/src/diffusers/pipelines/easyanimate/pipeline_easyanimate.py
@@ -96,7 +96,7 @@ def get_resize_crop_region_for_grid(src, tgt_width, tgt_height):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -123,7 +123,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -400,7 +400,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, prompt_attention_mask, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/easyanimate/pipeline_easyanimate_control.py b/src/diffusers/pipelines/easyanimate/pipeline_easyanimate_control.py
index f74a11f87d75..9fff4dc99c01 100755
--- a/src/diffusers/pipelines/easyanimate/pipeline_easyanimate_control.py
+++ b/src/diffusers/pipelines/easyanimate/pipeline_easyanimate_control.py
@@ -172,7 +172,7 @@ def get_resize_crop_region_for_grid(src, tgt_width, tgt_height):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -225,7 +225,7 @@ def resize_mask(mask, latent, process_first_frame_only=True):
     return resized_mask
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -511,7 +511,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, prompt_attention_mask, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/easyanimate/pipeline_easyanimate_inpaint.py b/src/diffusers/pipelines/easyanimate/pipeline_easyanimate_inpaint.py
index b16ef92d8e6b..f1a0b899a90b 100755
--- a/src/diffusers/pipelines/easyanimate/pipeline_easyanimate_inpaint.py
+++ b/src/diffusers/pipelines/easyanimate/pipeline_easyanimate_inpaint.py
@@ -194,7 +194,7 @@ def get_resize_crop_region_for_grid(src, tgt_width, tgt_height):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -267,7 +267,7 @@ def add_noise_to_reference_video(image, ratio=None, generator=None):
     return image
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -553,7 +553,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, prompt_attention_mask, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/flux/pipeline_flux.py b/src/diffusers/pipelines/flux/pipeline_flux.py
index 5041e352f73d..8d03ed9b270a 100644
--- a/src/diffusers/pipelines/flux/pipeline_flux.py
+++ b/src/diffusers/pipelines/flux/pipeline_flux.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -31,16 +30,13 @@
 from ...models import AutoencoderKL, FluxTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
 from ...utils import (
-    USE_PEFT_BACKEND,
-    deprecate,
     is_torch_xla_available,
     logging,
     replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
 )
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
+from .pipeline_flux_utils import FluxMixin, calculate_shift, retrieve_timesteps
 from .pipeline_output import FluxPipelineOutput
 
 
@@ -71,81 +67,9 @@
 """
 
 
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
 class FluxPipeline(
     DiffusionPipeline,
+    FluxMixin,
     FluxLoraLoaderMixin,
     FromSingleFileMixin,
     TextualInversionLoaderMixin,
@@ -215,178 +139,6 @@ def __init__(
         )
         self.default_sample_size = 128
 
-    def _get_t5_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        num_images_per_prompt: int = 1,
-        max_sequence_length: int = 512,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer_2)
-
-        text_inputs = self.tokenizer_2(
-            prompt,
-            padding="max_length",
-            max_length=max_sequence_length,
-            truncation=True,
-            return_length=False,
-            return_overflowing_tokens=False,
-            return_tensors="pt",
-        )
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer_2(prompt, padding="longest", return_tensors="pt").input_ids
-
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer_2.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because `max_sequence_length` is set to "
-                f" {max_sequence_length} tokens: {removed_text}"
-            )
-
-        prompt_embeds = self.text_encoder_2(text_input_ids.to(device), output_hidden_states=False)[0]
-
-        dtype = self.text_encoder_2.dtype
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        _, seq_len, _ = prompt_embeds.shape
-
-        # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        return prompt_embeds
-
-    def _get_clip_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]],
-        num_images_per_prompt: int = 1,
-        device: Optional[torch.device] = None,
-    ):
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-        text_inputs = self.tokenizer(
-            prompt,
-            padding="max_length",
-            max_length=self.tokenizer_max_length,
-            truncation=True,
-            return_overflowing_tokens=False,
-            return_length=False,
-            return_tensors="pt",
-        )
-
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because CLIP can only handle sequences up to"
-                f" {self.tokenizer_max_length} tokens: {removed_text}"
-            )
-        prompt_embeds = self.text_encoder(text_input_ids.to(device), output_hidden_states=False)
-
-        # Use pooled output of CLIPTextModel
-        prompt_embeds = prompt_embeds.pooler_output
-        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
-
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)
-
-        return prompt_embeds
-
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        prompt_2: Optional[Union[str, List[str]]] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
-        max_sequence_length: int = 512,
-        lora_scale: Optional[float] = None,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                used in all text-encoders
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-        """
-        device = device or self._execution_device
-
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, FluxLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder, lora_scale)
-            if self.text_encoder_2 is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder_2, lora_scale)
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        if prompt_embeds is None:
-            prompt_2 = prompt_2 or prompt
-            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
-
-            # We only use the pooled prompt output from the CLIPTextModel
-            pooled_prompt_embeds = self._get_clip_prompt_embeds(
-                prompt=prompt,
-                device=device,
-                num_images_per_prompt=num_images_per_prompt,
-            )
-            prompt_embeds = self._get_t5_prompt_embeds(
-                prompt=prompt_2,
-                num_images_per_prompt=num_images_per_prompt,
-                max_sequence_length=max_sequence_length,
-                device=device,
-            )
-
-        if self.text_encoder is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        if self.text_encoder_2 is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder_2, lora_scale)
-
-        dtype = self.text_encoder.dtype if self.text_encoder is not None else self.transformer.dtype
-        text_ids = torch.zeros(prompt_embeds.shape[1], 3).to(device=device, dtype=dtype)
-
-        return prompt_embeds, pooled_prompt_embeds, text_ids
-
     def encode_image(self, image, device, num_images_per_prompt):
         dtype = next(self.image_encoder.parameters()).dtype
 
@@ -503,97 +255,6 @@ def check_inputs(
         if max_sequence_length is not None and max_sequence_length > 512:
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
-    @staticmethod
-    def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
-        latent_image_ids = torch.zeros(height, width, 3)
-        latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
-        latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width)[None, :]
-
-        latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
-
-        latent_image_ids = latent_image_ids.reshape(
-            latent_image_id_height * latent_image_id_width, latent_image_id_channels
-        )
-
-        return latent_image_ids.to(device=device, dtype=dtype)
-
-    @staticmethod
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), height, width)
-
-        return latents
-
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
-        deprecate(
-            "enable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_slicing()
-
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
-        deprecate(
-            "disable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_slicing()
-
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
-        deprecate(
-            "enable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_tiling()
-
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
-        deprecate(
-            "disable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_tiling()
-
     def prepare_latents(
         self,
         batch_size,
@@ -629,26 +290,6 @@ def prepare_latents(
 
         return latents, latent_image_ids
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def joint_attention_kwargs(self):
-        return self._joint_attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def current_timestep(self):
-        return self._current_timestep
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/flux/pipeline_flux_control.py b/src/diffusers/pipelines/flux/pipeline_flux_control.py
index 848d7bd39254..dbeb0bce9e27 100644
--- a/src/diffusers/pipelines/flux/pipeline_flux_control.py
+++ b/src/diffusers/pipelines/flux/pipeline_flux_control.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -25,16 +24,13 @@
 from ...models.transformers import FluxTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
 from ...utils import (
-    USE_PEFT_BACKEND,
-    deprecate,
     is_torch_xla_available,
     logging,
     replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
 )
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
+from .pipeline_flux_utils import FluxControlMixin, calculate_shift, retrieve_timesteps
 from .pipeline_output import FluxPipelineOutput
 
 
@@ -83,82 +79,9 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
 class FluxControlPipeline(
     DiffusionPipeline,
+    FluxControlMixin,
     FluxLoraLoaderMixin,
     FromSingleFileMixin,
     TextualInversionLoaderMixin,
@@ -226,181 +149,6 @@ def __init__(
         )
         self.default_sample_size = 128
 
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_t5_prompt_embeds
-    def _get_t5_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        num_images_per_prompt: int = 1,
-        max_sequence_length: int = 512,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer_2)
-
-        text_inputs = self.tokenizer_2(
-            prompt,
-            padding="max_length",
-            max_length=max_sequence_length,
-            truncation=True,
-            return_length=False,
-            return_overflowing_tokens=False,
-            return_tensors="pt",
-        )
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer_2(prompt, padding="longest", return_tensors="pt").input_ids
-
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer_2.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because `max_sequence_length` is set to "
-                f" {max_sequence_length} tokens: {removed_text}"
-            )
-
-        prompt_embeds = self.text_encoder_2(text_input_ids.to(device), output_hidden_states=False)[0]
-
-        dtype = self.text_encoder_2.dtype
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        _, seq_len, _ = prompt_embeds.shape
-
-        # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_clip_prompt_embeds
-    def _get_clip_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]],
-        num_images_per_prompt: int = 1,
-        device: Optional[torch.device] = None,
-    ):
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-        text_inputs = self.tokenizer(
-            prompt,
-            padding="max_length",
-            max_length=self.tokenizer_max_length,
-            truncation=True,
-            return_overflowing_tokens=False,
-            return_length=False,
-            return_tensors="pt",
-        )
-
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because CLIP can only handle sequences up to"
-                f" {self.tokenizer_max_length} tokens: {removed_text}"
-            )
-        prompt_embeds = self.text_encoder(text_input_ids.to(device), output_hidden_states=False)
-
-        # Use pooled output of CLIPTextModel
-        prompt_embeds = prompt_embeds.pooler_output
-        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
-
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        prompt_2: Optional[Union[str, List[str]]] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
-        max_sequence_length: int = 512,
-        lora_scale: Optional[float] = None,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                used in all text-encoders
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-        """
-        device = device or self._execution_device
-
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, FluxLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder, lora_scale)
-            if self.text_encoder_2 is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder_2, lora_scale)
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        if prompt_embeds is None:
-            prompt_2 = prompt_2 or prompt
-            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
-
-            # We only use the pooled prompt output from the CLIPTextModel
-            pooled_prompt_embeds = self._get_clip_prompt_embeds(
-                prompt=prompt,
-                device=device,
-                num_images_per_prompt=num_images_per_prompt,
-            )
-            prompt_embeds = self._get_t5_prompt_embeds(
-                prompt=prompt_2,
-                num_images_per_prompt=num_images_per_prompt,
-                max_sequence_length=max_sequence_length,
-                device=device,
-            )
-
-        if self.text_encoder is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        if self.text_encoder_2 is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder_2, lora_scale)
-
-        dtype = self.text_encoder.dtype if self.text_encoder is not None else self.transformer.dtype
-        text_ids = torch.zeros(prompt_embeds.shape[1], 3).to(device=device, dtype=dtype)
-
-        return prompt_embeds, pooled_prompt_embeds, text_ids
-
     def check_inputs(
         self,
         prompt,
@@ -451,100 +199,6 @@ def check_inputs(
         if max_sequence_length is not None and max_sequence_length > 512:
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
-    def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
-        latent_image_ids = torch.zeros(height, width, 3)
-        latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
-        latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width)[None, :]
-
-        latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
-
-        latent_image_ids = latent_image_ids.reshape(
-            latent_image_id_height * latent_image_id_width, latent_image_id_channels
-        )
-
-        return latent_image_ids.to(device=device, dtype=dtype)
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._unpack_latents
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), height, width)
-
-        return latents
-
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
-        deprecate(
-            "enable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_slicing()
-
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
-        deprecate(
-            "disable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_slicing()
-
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
-        deprecate(
-            "enable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_tiling()
-
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
-        deprecate(
-            "disable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_tiling()
-
     # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.prepare_latents
     def prepare_latents(
         self,
@@ -581,57 +235,6 @@ def prepare_latents(
 
         return latents, latent_image_ids
 
-    # Copied from diffusers.pipelines.controlnet_sd3.pipeline_stable_diffusion_3_controlnet.StableDiffusion3ControlNetPipeline.prepare_image
-    def prepare_image(
-        self,
-        image,
-        width,
-        height,
-        batch_size,
-        num_images_per_prompt,
-        device,
-        dtype,
-        do_classifier_free_guidance=False,
-        guess_mode=False,
-    ):
-        if isinstance(image, torch.Tensor):
-            pass
-        else:
-            image = self.image_processor.preprocess(image, height=height, width=width)
-
-        image_batch_size = image.shape[0]
-
-        if image_batch_size == 1:
-            repeat_by = batch_size
-        else:
-            # image batch size is the same as prompt batch size
-            repeat_by = num_images_per_prompt
-
-        image = image.repeat_interleave(repeat_by, dim=0)
-
-        image = image.to(device=device, dtype=dtype)
-
-        if do_classifier_free_guidance and not guess_mode:
-            image = torch.cat([image] * 2)
-
-        return image
-
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def joint_attention_kwargs(self):
-        return self._joint_attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/flux/pipeline_flux_control_img2img.py b/src/diffusers/pipelines/flux/pipeline_flux_control_img2img.py
index 262345c75afc..f0269f5eca63 100644
--- a/src/diffusers/pipelines/flux/pipeline_flux_control_img2img.py
+++ b/src/diffusers/pipelines/flux/pipeline_flux_control_img2img.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -20,20 +19,14 @@
 from transformers import CLIPTextModel, CLIPTokenizer, T5EncoderModel, T5TokenizerFast
 
 from ...image_processor import PipelineImageInput, VaeImageProcessor
-from ...loaders import FluxLoraLoaderMixin, FromSingleFileMixin, TextualInversionLoaderMixin
+from ...loaders import FluxLoraLoaderMixin, FromSingleFileMixin
 from ...models.autoencoders import AutoencoderKL
 from ...models.transformers import FluxTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import (
-    USE_PEFT_BACKEND,
-    is_torch_xla_available,
-    logging,
-    replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
-)
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
+from .pipeline_flux_utils import FluxControlMixin, calculate_shift, retrieve_latents, retrieve_timesteps
 from .pipeline_output import FluxPipelineOutput
 
 
@@ -87,95 +80,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
-class FluxControlImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFileMixin):
+class FluxControlImg2ImgPipeline(DiffusionPipeline, FluxControlMixin, FluxLoraLoaderMixin, FromSingleFileMixin):
     r"""
     The Flux pipeline for image inpainting.
 
@@ -236,181 +141,6 @@ def __init__(
         )
         self.default_sample_size = 128
 
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_t5_prompt_embeds
-    def _get_t5_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        num_images_per_prompt: int = 1,
-        max_sequence_length: int = 512,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer_2)
-
-        text_inputs = self.tokenizer_2(
-            prompt,
-            padding="max_length",
-            max_length=max_sequence_length,
-            truncation=True,
-            return_length=False,
-            return_overflowing_tokens=False,
-            return_tensors="pt",
-        )
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer_2(prompt, padding="longest", return_tensors="pt").input_ids
-
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer_2.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because `max_sequence_length` is set to "
-                f" {max_sequence_length} tokens: {removed_text}"
-            )
-
-        prompt_embeds = self.text_encoder_2(text_input_ids.to(device), output_hidden_states=False)[0]
-
-        dtype = self.text_encoder_2.dtype
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        _, seq_len, _ = prompt_embeds.shape
-
-        # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_clip_prompt_embeds
-    def _get_clip_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]],
-        num_images_per_prompt: int = 1,
-        device: Optional[torch.device] = None,
-    ):
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-        text_inputs = self.tokenizer(
-            prompt,
-            padding="max_length",
-            max_length=self.tokenizer_max_length,
-            truncation=True,
-            return_overflowing_tokens=False,
-            return_length=False,
-            return_tensors="pt",
-        )
-
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because CLIP can only handle sequences up to"
-                f" {self.tokenizer_max_length} tokens: {removed_text}"
-            )
-        prompt_embeds = self.text_encoder(text_input_ids.to(device), output_hidden_states=False)
-
-        # Use pooled output of CLIPTextModel
-        prompt_embeds = prompt_embeds.pooler_output
-        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
-
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        prompt_2: Optional[Union[str, List[str]]] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
-        max_sequence_length: int = 512,
-        lora_scale: Optional[float] = None,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                used in all text-encoders
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-        """
-        device = device or self._execution_device
-
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, FluxLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder, lora_scale)
-            if self.text_encoder_2 is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder_2, lora_scale)
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        if prompt_embeds is None:
-            prompt_2 = prompt_2 or prompt
-            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
-
-            # We only use the pooled prompt output from the CLIPTextModel
-            pooled_prompt_embeds = self._get_clip_prompt_embeds(
-                prompt=prompt,
-                device=device,
-                num_images_per_prompt=num_images_per_prompt,
-            )
-            prompt_embeds = self._get_t5_prompt_embeds(
-                prompt=prompt_2,
-                num_images_per_prompt=num_images_per_prompt,
-                max_sequence_length=max_sequence_length,
-                device=device,
-            )
-
-        if self.text_encoder is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        if self.text_encoder_2 is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder_2, lora_scale)
-
-        dtype = self.text_encoder.dtype if self.text_encoder is not None else self.transformer.dtype
-        text_ids = torch.zeros(prompt_embeds.shape[1], 3).to(device=device, dtype=dtype)
-
-        return prompt_embeds, pooled_prompt_embeds, text_ids
-
     # Copied from diffusers.pipelines.stable_diffusion_3.pipeline_stable_diffusion_3_inpaint.StableDiffusion3InpaintPipeline._encode_vae_image
     def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
         if isinstance(generator, list):
@@ -492,47 +222,6 @@ def check_inputs(
         if max_sequence_length is not None and max_sequence_length > 512:
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
-    def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
-        latent_image_ids = torch.zeros(height, width, 3)
-        latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
-        latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width)[None, :]
-
-        latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
-
-        latent_image_ids = latent_image_ids.reshape(
-            latent_image_id_height * latent_image_id_width, latent_image_id_channels
-        )
-
-        return latent_image_ids.to(device=device, dtype=dtype)
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._unpack_latents
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), height, width)
-
-        return latents
-
     def prepare_latents(
         self,
         image,
@@ -615,22 +304,6 @@ def prepare_image(
 
         return image
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def joint_attention_kwargs(self):
-        return self._joint_attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/flux/pipeline_flux_control_inpaint.py b/src/diffusers/pipelines/flux/pipeline_flux_control_inpaint.py
index 6915a83a7ca7..d475c54f73cf 100644
--- a/src/diffusers/pipelines/flux/pipeline_flux_control_inpaint.py
+++ b/src/diffusers/pipelines/flux/pipeline_flux_control_inpaint.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -33,17 +32,10 @@
 from ...models.autoencoders import AutoencoderKL
 from ...models.transformers import FluxTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import (
-    USE_PEFT_BACKEND,
-    deprecate,
-    is_torch_xla_available,
-    logging,
-    replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
-)
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
+from .pipeline_flux_utils import FluxControlMixin, calculate_shift, retrieve_latents, retrieve_timesteps
 from .pipeline_output import FluxPipelineOutput
 
 
@@ -114,96 +106,9 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
 class FluxControlInpaintPipeline(
     DiffusionPipeline,
+    FluxControlMixin,
     FluxLoraLoaderMixin,
     FromSingleFileMixin,
     TextualInversionLoaderMixin,
@@ -276,181 +181,6 @@ def __init__(
         )
         self.default_sample_size = 128
 
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_t5_prompt_embeds
-    def _get_t5_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        num_images_per_prompt: int = 1,
-        max_sequence_length: int = 512,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer_2)
-
-        text_inputs = self.tokenizer_2(
-            prompt,
-            padding="max_length",
-            max_length=max_sequence_length,
-            truncation=True,
-            return_length=False,
-            return_overflowing_tokens=False,
-            return_tensors="pt",
-        )
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer_2(prompt, padding="longest", return_tensors="pt").input_ids
-
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer_2.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because `max_sequence_length` is set to "
-                f" {max_sequence_length} tokens: {removed_text}"
-            )
-
-        prompt_embeds = self.text_encoder_2(text_input_ids.to(device), output_hidden_states=False)[0]
-
-        dtype = self.text_encoder_2.dtype
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        _, seq_len, _ = prompt_embeds.shape
-
-        # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_clip_prompt_embeds
-    def _get_clip_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]],
-        num_images_per_prompt: int = 1,
-        device: Optional[torch.device] = None,
-    ):
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-        text_inputs = self.tokenizer(
-            prompt,
-            padding="max_length",
-            max_length=self.tokenizer_max_length,
-            truncation=True,
-            return_overflowing_tokens=False,
-            return_length=False,
-            return_tensors="pt",
-        )
-
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because CLIP can only handle sequences up to"
-                f" {self.tokenizer_max_length} tokens: {removed_text}"
-            )
-        prompt_embeds = self.text_encoder(text_input_ids.to(device), output_hidden_states=False)
-
-        # Use pooled output of CLIPTextModel
-        prompt_embeds = prompt_embeds.pooler_output
-        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
-
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        prompt_2: Optional[Union[str, List[str]]] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
-        max_sequence_length: int = 512,
-        lora_scale: Optional[float] = None,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                used in all text-encoders
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-        """
-        device = device or self._execution_device
-
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, FluxLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder, lora_scale)
-            if self.text_encoder_2 is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder_2, lora_scale)
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        if prompt_embeds is None:
-            prompt_2 = prompt_2 or prompt
-            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
-
-            # We only use the pooled prompt output from the CLIPTextModel
-            pooled_prompt_embeds = self._get_clip_prompt_embeds(
-                prompt=prompt,
-                device=device,
-                num_images_per_prompt=num_images_per_prompt,
-            )
-            prompt_embeds = self._get_t5_prompt_embeds(
-                prompt=prompt_2,
-                num_images_per_prompt=num_images_per_prompt,
-                max_sequence_length=max_sequence_length,
-                device=device,
-            )
-
-        if self.text_encoder is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        if self.text_encoder_2 is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder_2, lora_scale)
-
-        dtype = self.text_encoder.dtype if self.text_encoder is not None else self.transformer.dtype
-        text_ids = torch.zeros(prompt_embeds.shape[1], 3).to(device=device, dtype=dtype)
-
-        return prompt_embeds, pooled_prompt_embeds, text_ids
-
     # Copied from diffusers.pipelines.stable_diffusion_3.pipeline_stable_diffusion_3_inpaint.StableDiffusion3InpaintPipeline._encode_vae_image
     def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
         if isinstance(generator, list):
@@ -532,100 +262,6 @@ def check_inputs(
         if max_sequence_length is not None and max_sequence_length > 512:
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
-    def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
-        latent_image_ids = torch.zeros(height, width, 3)
-        latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
-        latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width)[None, :]
-
-        latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
-
-        latent_image_ids = latent_image_ids.reshape(
-            latent_image_id_height * latent_image_id_width, latent_image_id_channels
-        )
-
-        return latent_image_ids.to(device=device, dtype=dtype)
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._unpack_latents
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), height, width)
-
-        return latents
-
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
-        deprecate(
-            "enable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_slicing()
-
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
-        deprecate(
-            "disable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_slicing()
-
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
-        deprecate(
-            "enable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_tiling()
-
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
-        deprecate(
-            "disable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_tiling()
-
     def prepare_latents(
         self,
         image,
@@ -786,22 +422,6 @@ def prepare_mask_latents(
 
         return mask_image, masked_image_latents
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def joint_attention_kwargs(self):
-        return self._joint_attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/flux/pipeline_flux_controlnet.py b/src/diffusers/pipelines/flux/pipeline_flux_controlnet.py
index 507ec687347c..9209a51e7f42 100644
--- a/src/diffusers/pipelines/flux/pipeline_flux_controlnet.py
+++ b/src/diffusers/pipelines/flux/pipeline_flux_controlnet.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Tuple, Union
 
 import numpy as np
@@ -27,21 +26,15 @@
 )
 
 from ...image_processor import PipelineImageInput, VaeImageProcessor
-from ...loaders import FluxIPAdapterMixin, FluxLoraLoaderMixin, FromSingleFileMixin, TextualInversionLoaderMixin
+from ...loaders import FluxIPAdapterMixin, FluxLoraLoaderMixin, FromSingleFileMixin
 from ...models.autoencoders import AutoencoderKL
 from ...models.controlnets.controlnet_flux import FluxControlNetModel, FluxMultiControlNetModel
 from ...models.transformers import FluxTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import (
-    USE_PEFT_BACKEND,
-    is_torch_xla_available,
-    logging,
-    replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
-)
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
+from .pipeline_flux_utils import FluxControlMixin, calculate_shift, retrieve_latents, retrieve_timesteps
 from .pipeline_output import FluxPipelineOutput
 
 
@@ -86,95 +79,9 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
+class FluxControlNetPipeline(
+    DiffusionPipeline, FluxControlMixin, FluxLoraLoaderMixin, FromSingleFileMixin, FluxIPAdapterMixin
 ):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
-class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFileMixin, FluxIPAdapterMixin):
     r"""
     The Flux pipeline for text-to-image generation.
 
@@ -245,181 +152,6 @@ def __init__(
         )
         self.default_sample_size = 128
 
-    def _get_t5_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        num_images_per_prompt: int = 1,
-        max_sequence_length: int = 512,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-        text_inputs = self.tokenizer_2(
-            prompt,
-            padding="max_length",
-            max_length=max_sequence_length,
-            truncation=True,
-            return_length=False,
-            return_overflowing_tokens=False,
-            return_tensors="pt",
-        )
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer_2(prompt, padding="longest", return_tensors="pt").input_ids
-
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer_2.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because `max_sequence_length` is set to "
-                f" {max_sequence_length} tokens: {removed_text}"
-            )
-
-        prompt_embeds = self.text_encoder_2(text_input_ids.to(device), output_hidden_states=False)[0]
-
-        dtype = self.text_encoder_2.dtype
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        _, seq_len, _ = prompt_embeds.shape
-
-        # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        return prompt_embeds
-
-    def _get_clip_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]],
-        num_images_per_prompt: int = 1,
-        device: Optional[torch.device] = None,
-    ):
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-        text_inputs = self.tokenizer(
-            prompt,
-            padding="max_length",
-            max_length=self.tokenizer_max_length,
-            truncation=True,
-            return_overflowing_tokens=False,
-            return_length=False,
-            return_tensors="pt",
-        )
-
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because CLIP can only handle sequences up to"
-                f" {self.tokenizer_max_length} tokens: {removed_text}"
-            )
-        prompt_embeds = self.text_encoder(text_input_ids.to(device), output_hidden_states=False)
-
-        # Use pooled output of CLIPTextModel
-        prompt_embeds = prompt_embeds.pooler_output
-        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
-
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)
-
-        return prompt_embeds
-
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        prompt_2: Optional[Union[str, List[str]]] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
-        max_sequence_length: int = 512,
-        lora_scale: Optional[float] = None,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                used in all text-encoders
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            clip_skip (`int`, *optional*):
-                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
-                the output of the pre-final layer will be used for computing the prompt embeddings.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-        """
-        device = device or self._execution_device
-
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, FluxLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder, lora_scale)
-            if self.text_encoder_2 is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder_2, lora_scale)
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        if prompt_embeds is None:
-            prompt_2 = prompt_2 or prompt
-            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
-
-            # We only use the pooled prompt output from the CLIPTextModel
-            pooled_prompt_embeds = self._get_clip_prompt_embeds(
-                prompt=prompt,
-                device=device,
-                num_images_per_prompt=num_images_per_prompt,
-            )
-            prompt_embeds = self._get_t5_prompt_embeds(
-                prompt=prompt_2,
-                num_images_per_prompt=num_images_per_prompt,
-                max_sequence_length=max_sequence_length,
-                device=device,
-            )
-
-        if self.text_encoder is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        if self.text_encoder_2 is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder_2, lora_scale)
-
-        dtype = self.text_encoder.dtype if self.text_encoder is not None else self.transformer.dtype
-        text_ids = torch.zeros(prompt_embeds.shape[1], 3).to(device=device, dtype=dtype)
-
-        return prompt_embeds, pooled_prompt_embeds, text_ids
-
     # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_image
     def encode_image(self, image, device, num_images_per_prompt):
         dtype = next(self.image_encoder.parameters()).dtype
@@ -546,47 +278,6 @@ def check_inputs(
         if max_sequence_length is not None and max_sequence_length > 512:
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
-    def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
-        latent_image_ids = torch.zeros(height, width, 3)
-        latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
-        latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width)[None, :]
-
-        latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
-
-        latent_image_ids = latent_image_ids.reshape(
-            latent_image_id_height * latent_image_id_width, latent_image_id_channels
-        )
-
-        return latent_image_ids.to(device=device, dtype=dtype)
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._unpack_latents
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), height, width)
-
-        return latents
-
     # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.prepare_latents
     def prepare_latents(
         self,
@@ -658,22 +349,6 @@ def prepare_image(
 
         return image
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def joint_attention_kwargs(self):
-        return self._joint_attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/flux/pipeline_flux_controlnet_image_to_image.py b/src/diffusers/pipelines/flux/pipeline_flux_controlnet_image_to_image.py
index 582c7bbad84e..0cd4a051ae45 100644
--- a/src/diffusers/pipelines/flux/pipeline_flux_controlnet_image_to_image.py
+++ b/src/diffusers/pipelines/flux/pipeline_flux_controlnet_image_to_image.py
@@ -1,4 +1,3 @@
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Tuple, Union
 
 import numpy as np
@@ -11,21 +10,15 @@
 )
 
 from ...image_processor import PipelineImageInput, VaeImageProcessor
-from ...loaders import FluxLoraLoaderMixin, FromSingleFileMixin, TextualInversionLoaderMixin
+from ...loaders import FluxLoraLoaderMixin, FromSingleFileMixin
 from ...models.autoencoders import AutoencoderKL
 from ...models.controlnets.controlnet_flux import FluxControlNetModel, FluxMultiControlNetModel
 from ...models.transformers import FluxTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import (
-    USE_PEFT_BACKEND,
-    is_torch_xla_available,
-    logging,
-    replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
-)
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
+from .pipeline_flux_utils import FluxControlMixin, calculate_shift, retrieve_latents, retrieve_timesteps
 from .pipeline_output import FluxPipelineOutput
 
 
@@ -81,95 +74,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
-class FluxControlNetImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFileMixin):
+class FluxControlNetImg2ImgPipeline(DiffusionPipeline, FluxControlMixin, FluxLoraLoaderMixin, FromSingleFileMixin):
     r"""
     The Flux controlnet pipeline for image-to-image generation.
 
@@ -236,181 +141,6 @@ def __init__(
         )
         self.default_sample_size = 128
 
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_t5_prompt_embeds
-    def _get_t5_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        num_images_per_prompt: int = 1,
-        max_sequence_length: int = 512,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer_2)
-
-        text_inputs = self.tokenizer_2(
-            prompt,
-            padding="max_length",
-            max_length=max_sequence_length,
-            truncation=True,
-            return_length=False,
-            return_overflowing_tokens=False,
-            return_tensors="pt",
-        )
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer_2(prompt, padding="longest", return_tensors="pt").input_ids
-
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer_2.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because `max_sequence_length` is set to "
-                f" {max_sequence_length} tokens: {removed_text}"
-            )
-
-        prompt_embeds = self.text_encoder_2(text_input_ids.to(device), output_hidden_states=False)[0]
-
-        dtype = self.text_encoder_2.dtype
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        _, seq_len, _ = prompt_embeds.shape
-
-        # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_clip_prompt_embeds
-    def _get_clip_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]],
-        num_images_per_prompt: int = 1,
-        device: Optional[torch.device] = None,
-    ):
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-        text_inputs = self.tokenizer(
-            prompt,
-            padding="max_length",
-            max_length=self.tokenizer_max_length,
-            truncation=True,
-            return_overflowing_tokens=False,
-            return_length=False,
-            return_tensors="pt",
-        )
-
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because CLIP can only handle sequences up to"
-                f" {self.tokenizer_max_length} tokens: {removed_text}"
-            )
-        prompt_embeds = self.text_encoder(text_input_ids.to(device), output_hidden_states=False)
-
-        # Use pooled output of CLIPTextModel
-        prompt_embeds = prompt_embeds.pooler_output
-        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
-
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        prompt_2: Optional[Union[str, List[str]]] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
-        max_sequence_length: int = 512,
-        lora_scale: Optional[float] = None,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                used in all text-encoders
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-        """
-        device = device or self._execution_device
-
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, FluxLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder, lora_scale)
-            if self.text_encoder_2 is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder_2, lora_scale)
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        if prompt_embeds is None:
-            prompt_2 = prompt_2 or prompt
-            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
-
-            # We only use the pooled prompt output from the CLIPTextModel
-            pooled_prompt_embeds = self._get_clip_prompt_embeds(
-                prompt=prompt,
-                device=device,
-                num_images_per_prompt=num_images_per_prompt,
-            )
-            prompt_embeds = self._get_t5_prompt_embeds(
-                prompt=prompt_2,
-                num_images_per_prompt=num_images_per_prompt,
-                max_sequence_length=max_sequence_length,
-                device=device,
-            )
-
-        if self.text_encoder is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        if self.text_encoder_2 is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder_2, lora_scale)
-
-        dtype = self.text_encoder.dtype if self.text_encoder is not None else self.transformer.dtype
-        text_ids = torch.zeros(prompt_embeds.shape[1], 3).to(device=device, dtype=dtype)
-
-        return prompt_embeds, pooled_prompt_embeds, text_ids
-
     # Copied from diffusers.pipelines.stable_diffusion_3.pipeline_stable_diffusion_3_inpaint.StableDiffusion3InpaintPipeline._encode_vae_image
     def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
         if isinstance(generator, list):
@@ -492,47 +222,6 @@ def check_inputs(
         if max_sequence_length is not None and max_sequence_length > 512:
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
-    def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
-        latent_image_ids = torch.zeros(height, width, 3)
-        latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
-        latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width)[None, :]
-
-        latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
-
-        latent_image_ids = latent_image_ids.reshape(
-            latent_image_id_height * latent_image_id_width, latent_image_id_channels
-        )
-
-        return latent_image_ids.to(device=device, dtype=dtype)
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._unpack_latents
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), height, width)
-
-        return latents
-
     def prepare_latents(
         self,
         image,
@@ -615,22 +304,6 @@ def prepare_image(
 
         return image
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def joint_attention_kwargs(self):
-        return self._joint_attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/flux/pipeline_flux_controlnet_inpainting.py b/src/diffusers/pipelines/flux/pipeline_flux_controlnet_inpainting.py
index f7f34ef231af..9fb7f278f306 100644
--- a/src/diffusers/pipelines/flux/pipeline_flux_controlnet_inpainting.py
+++ b/src/diffusers/pipelines/flux/pipeline_flux_controlnet_inpainting.py
@@ -1,4 +1,3 @@
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Tuple, Union
 
 import numpy as np
@@ -12,21 +11,15 @@
 )
 
 from ...image_processor import PipelineImageInput, VaeImageProcessor
-from ...loaders import FluxLoraLoaderMixin, FromSingleFileMixin, TextualInversionLoaderMixin
+from ...loaders import FluxLoraLoaderMixin, FromSingleFileMixin
 from ...models.autoencoders import AutoencoderKL
 from ...models.controlnets.controlnet_flux import FluxControlNetModel, FluxMultiControlNetModel
 from ...models.transformers import FluxTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import (
-    USE_PEFT_BACKEND,
-    is_torch_xla_available,
-    logging,
-    replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
-)
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
+from .pipeline_flux_utils import FluxControlMixin, calculate_shift, retrieve_latents, retrieve_timesteps
 from .pipeline_output import FluxPipelineOutput
 
 
@@ -83,95 +76,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
-class FluxControlNetInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFileMixin):
+class FluxControlNetInpaintPipeline(DiffusionPipeline, FluxControlMixin, FluxLoraLoaderMixin, FromSingleFileMixin):
     r"""
     The Flux controlnet pipeline for inpainting.
 
@@ -247,181 +152,6 @@ def __init__(
         )
         self.default_sample_size = 128
 
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_t5_prompt_embeds
-    def _get_t5_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        num_images_per_prompt: int = 1,
-        max_sequence_length: int = 512,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer_2)
-
-        text_inputs = self.tokenizer_2(
-            prompt,
-            padding="max_length",
-            max_length=max_sequence_length,
-            truncation=True,
-            return_length=False,
-            return_overflowing_tokens=False,
-            return_tensors="pt",
-        )
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer_2(prompt, padding="longest", return_tensors="pt").input_ids
-
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer_2.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because `max_sequence_length` is set to "
-                f" {max_sequence_length} tokens: {removed_text}"
-            )
-
-        prompt_embeds = self.text_encoder_2(text_input_ids.to(device), output_hidden_states=False)[0]
-
-        dtype = self.text_encoder_2.dtype
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        _, seq_len, _ = prompt_embeds.shape
-
-        # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_clip_prompt_embeds
-    def _get_clip_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]],
-        num_images_per_prompt: int = 1,
-        device: Optional[torch.device] = None,
-    ):
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-        text_inputs = self.tokenizer(
-            prompt,
-            padding="max_length",
-            max_length=self.tokenizer_max_length,
-            truncation=True,
-            return_overflowing_tokens=False,
-            return_length=False,
-            return_tensors="pt",
-        )
-
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because CLIP can only handle sequences up to"
-                f" {self.tokenizer_max_length} tokens: {removed_text}"
-            )
-        prompt_embeds = self.text_encoder(text_input_ids.to(device), output_hidden_states=False)
-
-        # Use pooled output of CLIPTextModel
-        prompt_embeds = prompt_embeds.pooler_output
-        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
-
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        prompt_2: Optional[Union[str, List[str]]] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
-        max_sequence_length: int = 512,
-        lora_scale: Optional[float] = None,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                used in all text-encoders
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-        """
-        device = device or self._execution_device
-
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, FluxLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder, lora_scale)
-            if self.text_encoder_2 is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder_2, lora_scale)
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        if prompt_embeds is None:
-            prompt_2 = prompt_2 or prompt
-            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
-
-            # We only use the pooled prompt output from the CLIPTextModel
-            pooled_prompt_embeds = self._get_clip_prompt_embeds(
-                prompt=prompt,
-                device=device,
-                num_images_per_prompt=num_images_per_prompt,
-            )
-            prompt_embeds = self._get_t5_prompt_embeds(
-                prompt=prompt_2,
-                num_images_per_prompt=num_images_per_prompt,
-                max_sequence_length=max_sequence_length,
-                device=device,
-            )
-
-        if self.text_encoder is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        if self.text_encoder_2 is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder_2, lora_scale)
-
-        dtype = self.text_encoder.dtype if self.text_encoder is not None else self.transformer.dtype
-        text_ids = torch.zeros(prompt_embeds.shape[1], 3).to(device=device, dtype=dtype)
-
-        return prompt_embeds, pooled_prompt_embeds, text_ids
-
     # Copied from diffusers.pipelines.stable_diffusion_3.pipeline_stable_diffusion_3_inpaint.StableDiffusion3InpaintPipeline._encode_vae_image
     def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
         if isinstance(generator, list):
@@ -520,47 +250,6 @@ def check_inputs(
         if max_sequence_length is not None and max_sequence_length > 512:
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
-    def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
-        latent_image_ids = torch.zeros(height, width, 3)
-        latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
-        latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width)[None, :]
-
-        latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
-
-        latent_image_ids = latent_image_ids.reshape(
-            latent_image_id_height * latent_image_id_width, latent_image_id_channels
-        )
-
-        return latent_image_ids.to(device=device, dtype=dtype)
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._unpack_latents
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), height, width)
-
-        return latents
-
     def prepare_latents(
         self,
         image,
@@ -719,22 +408,6 @@ def prepare_image(
 
         return image
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def joint_attention_kwargs(self):
-        return self._joint_attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/flux/pipeline_flux_fill.py b/src/diffusers/pipelines/flux/pipeline_flux_fill.py
index 5cb9c82204b2..aadf8e607702 100644
--- a/src/diffusers/pipelines/flux/pipeline_flux_fill.py
+++ b/src/diffusers/pipelines/flux/pipeline_flux_fill.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -24,17 +23,10 @@
 from ...models.autoencoders import AutoencoderKL
 from ...models.transformers import FluxTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import (
-    USE_PEFT_BACKEND,
-    deprecate,
-    is_torch_xla_available,
-    logging,
-    replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
-)
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
+from .pipeline_flux_utils import FluxMixin, calculate_shift, retrieve_latents, retrieve_timesteps
 from .pipeline_output import FluxPipelineOutput
 
 
@@ -77,96 +69,9 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
 class FluxFillPipeline(
     DiffusionPipeline,
+    FluxMixin,
     FluxLoraLoaderMixin,
     FromSingleFileMixin,
     TextualInversionLoaderMixin,
@@ -241,101 +146,6 @@ def __init__(
         )
         self.default_sample_size = 128
 
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_t5_prompt_embeds
-    def _get_t5_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        num_images_per_prompt: int = 1,
-        max_sequence_length: int = 512,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer_2)
-
-        text_inputs = self.tokenizer_2(
-            prompt,
-            padding="max_length",
-            max_length=max_sequence_length,
-            truncation=True,
-            return_length=False,
-            return_overflowing_tokens=False,
-            return_tensors="pt",
-        )
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer_2(prompt, padding="longest", return_tensors="pt").input_ids
-
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer_2.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because `max_sequence_length` is set to "
-                f" {max_sequence_length} tokens: {removed_text}"
-            )
-
-        prompt_embeds = self.text_encoder_2(text_input_ids.to(device), output_hidden_states=False)[0]
-
-        dtype = self.text_encoder_2.dtype
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        _, seq_len, _ = prompt_embeds.shape
-
-        # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_clip_prompt_embeds
-    def _get_clip_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]],
-        num_images_per_prompt: int = 1,
-        device: Optional[torch.device] = None,
-    ):
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-        text_inputs = self.tokenizer(
-            prompt,
-            padding="max_length",
-            max_length=self.tokenizer_max_length,
-            truncation=True,
-            return_overflowing_tokens=False,
-            return_length=False,
-            return_tensors="pt",
-        )
-
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because CLIP can only handle sequences up to"
-                f" {self.tokenizer_max_length} tokens: {removed_text}"
-            )
-        prompt_embeds = self.text_encoder(text_input_ids.to(device), output_hidden_states=False)
-
-        # Use pooled output of CLIPTextModel
-        prompt_embeds = prompt_embeds.pooler_output
-        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
-
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)
-
-        return prompt_embeds
-
     def prepare_mask_latents(
         self,
         mask,
@@ -416,86 +226,6 @@ def prepare_mask_latents(
 
         return mask, masked_image_latents
 
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        prompt_2: Optional[Union[str, List[str]]] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
-        max_sequence_length: int = 512,
-        lora_scale: Optional[float] = None,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                used in all text-encoders
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-        """
-        device = device or self._execution_device
-
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, FluxLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder, lora_scale)
-            if self.text_encoder_2 is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder_2, lora_scale)
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        if prompt_embeds is None:
-            prompt_2 = prompt_2 or prompt
-            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
-
-            # We only use the pooled prompt output from the CLIPTextModel
-            pooled_prompt_embeds = self._get_clip_prompt_embeds(
-                prompt=prompt,
-                device=device,
-                num_images_per_prompt=num_images_per_prompt,
-            )
-            prompt_embeds = self._get_t5_prompt_embeds(
-                prompt=prompt_2,
-                num_images_per_prompt=num_images_per_prompt,
-                max_sequence_length=max_sequence_length,
-                device=device,
-            )
-
-        if self.text_encoder is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        if self.text_encoder_2 is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder_2, lora_scale)
-
-        dtype = self.text_encoder.dtype if self.text_encoder is not None else self.transformer.dtype
-        text_ids = torch.zeros(prompt_embeds.shape[1], 3).to(device=device, dtype=dtype)
-
-        return prompt_embeds, pooled_prompt_embeds, text_ids
-
     # Copied from diffusers.pipelines.stable_diffusion_3.pipeline_stable_diffusion_3_inpaint.StableDiffusion3InpaintPipeline._encode_vae_image
     def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
         if isinstance(generator, list):
@@ -588,100 +318,6 @@ def check_inputs(
         if image is not None and mask_image is None:
             raise ValueError("Please provide `mask_image` when passing `image`.")
 
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
-    def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
-        latent_image_ids = torch.zeros(height, width, 3)
-        latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
-        latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width)[None, :]
-
-        latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
-
-        latent_image_ids = latent_image_ids.reshape(
-            latent_image_id_height * latent_image_id_width, latent_image_id_channels
-        )
-
-        return latent_image_ids.to(device=device, dtype=dtype)
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._unpack_latents
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), height, width)
-
-        return latents
-
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
-        deprecate(
-            "enable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_slicing()
-
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
-        deprecate(
-            "disable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_slicing()
-
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
-        deprecate(
-            "enable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_tiling()
-
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
-        deprecate(
-            "disable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_tiling()
-
     # Copied from diffusers.pipelines.flux.pipeline_flux_img2img.FluxImg2ImgPipeline.prepare_latents
     def prepare_latents(
         self,
@@ -733,22 +369,6 @@ def prepare_latents(
         latents = self._pack_latents(latents, batch_size, num_channels_latents, height, width)
         return latents, latent_image_ids
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def joint_attention_kwargs(self):
-        return self._joint_attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/flux/pipeline_flux_img2img.py b/src/diffusers/pipelines/flux/pipeline_flux_img2img.py
index ab9140dae921..f8ba95088378 100644
--- a/src/diffusers/pipelines/flux/pipeline_flux_img2img.py
+++ b/src/diffusers/pipelines/flux/pipeline_flux_img2img.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -27,21 +26,14 @@
 )
 
 from ...image_processor import PipelineImageInput, VaeImageProcessor
-from ...loaders import FluxIPAdapterMixin, FluxLoraLoaderMixin, FromSingleFileMixin, TextualInversionLoaderMixin
+from ...loaders import FluxIPAdapterMixin, FluxLoraLoaderMixin, FromSingleFileMixin
 from ...models.autoencoders import AutoencoderKL
 from ...models.transformers import FluxTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import (
-    USE_PEFT_BACKEND,
-    deprecate,
-    is_torch_xla_available,
-    logging,
-    replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
-)
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
+from .pipeline_flux_utils import FluxMixin, calculate_shift, retrieve_latents, retrieve_timesteps
 from .pipeline_output import FluxPipelineOutput
 
 
@@ -79,95 +71,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
-class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFileMixin, FluxIPAdapterMixin):
+class FluxImg2ImgPipeline(DiffusionPipeline, FluxMixin, FluxLoraLoaderMixin, FromSingleFileMixin, FluxIPAdapterMixin):
     r"""
     The Flux pipeline for image inpainting.
 
@@ -235,181 +139,6 @@ def __init__(
         )
         self.default_sample_size = 128
 
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_t5_prompt_embeds
-    def _get_t5_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        num_images_per_prompt: int = 1,
-        max_sequence_length: int = 512,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer_2)
-
-        text_inputs = self.tokenizer_2(
-            prompt,
-            padding="max_length",
-            max_length=max_sequence_length,
-            truncation=True,
-            return_length=False,
-            return_overflowing_tokens=False,
-            return_tensors="pt",
-        )
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer_2(prompt, padding="longest", return_tensors="pt").input_ids
-
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer_2.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because `max_sequence_length` is set to "
-                f" {max_sequence_length} tokens: {removed_text}"
-            )
-
-        prompt_embeds = self.text_encoder_2(text_input_ids.to(device), output_hidden_states=False)[0]
-
-        dtype = self.text_encoder_2.dtype
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        _, seq_len, _ = prompt_embeds.shape
-
-        # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_clip_prompt_embeds
-    def _get_clip_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]],
-        num_images_per_prompt: int = 1,
-        device: Optional[torch.device] = None,
-    ):
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-        text_inputs = self.tokenizer(
-            prompt,
-            padding="max_length",
-            max_length=self.tokenizer_max_length,
-            truncation=True,
-            return_overflowing_tokens=False,
-            return_length=False,
-            return_tensors="pt",
-        )
-
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because CLIP can only handle sequences up to"
-                f" {self.tokenizer_max_length} tokens: {removed_text}"
-            )
-        prompt_embeds = self.text_encoder(text_input_ids.to(device), output_hidden_states=False)
-
-        # Use pooled output of CLIPTextModel
-        prompt_embeds = prompt_embeds.pooler_output
-        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
-
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        prompt_2: Optional[Union[str, List[str]]] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
-        max_sequence_length: int = 512,
-        lora_scale: Optional[float] = None,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                used in all text-encoders
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-        """
-        device = device or self._execution_device
-
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, FluxLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder, lora_scale)
-            if self.text_encoder_2 is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder_2, lora_scale)
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        if prompt_embeds is None:
-            prompt_2 = prompt_2 or prompt
-            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
-
-            # We only use the pooled prompt output from the CLIPTextModel
-            pooled_prompt_embeds = self._get_clip_prompt_embeds(
-                prompt=prompt,
-                device=device,
-                num_images_per_prompt=num_images_per_prompt,
-            )
-            prompt_embeds = self._get_t5_prompt_embeds(
-                prompt=prompt_2,
-                num_images_per_prompt=num_images_per_prompt,
-                max_sequence_length=max_sequence_length,
-                device=device,
-            )
-
-        if self.text_encoder is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        if self.text_encoder_2 is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder_2, lora_scale)
-
-        dtype = self.text_encoder.dtype if self.text_encoder is not None else self.transformer.dtype
-        text_ids = torch.zeros(prompt_embeds.shape[1], 3).to(device=device, dtype=dtype)
-
-        return prompt_embeds, pooled_prompt_embeds, text_ids
-
     # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_image
     def encode_image(self, image, device, num_images_per_prompt):
         dtype = next(self.image_encoder.parameters()).dtype
@@ -567,104 +296,6 @@ def check_inputs(
         if max_sequence_length is not None and max_sequence_length > 512:
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
-    def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
-        latent_image_ids = torch.zeros(height, width, 3)
-        latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
-        latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width)[None, :]
-
-        latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
-
-        latent_image_ids = latent_image_ids.reshape(
-            latent_image_id_height * latent_image_id_width, latent_image_id_channels
-        )
-
-        return latent_image_ids.to(device=device, dtype=dtype)
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._unpack_latents
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), height, width)
-
-        return latents
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.enable_vae_slicing
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
-        deprecate(
-            "enable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_slicing()
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.disable_vae_slicing
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
-        deprecate(
-            "disable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_slicing()
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.enable_vae_tiling
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
-        deprecate(
-            "enable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_tiling()
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.disable_vae_tiling
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
-        deprecate(
-            "disable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_tiling()
-
     def prepare_latents(
         self,
         image,
@@ -715,22 +346,6 @@ def prepare_latents(
         latents = self._pack_latents(latents, batch_size, num_channels_latents, height, width)
         return latents, latent_image_ids
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def joint_attention_kwargs(self):
-        return self._joint_attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/flux/pipeline_flux_inpaint.py b/src/diffusers/pipelines/flux/pipeline_flux_inpaint.py
index 3bfe82cf4382..cedbbd3953fc 100644
--- a/src/diffusers/pipelines/flux/pipeline_flux_inpaint.py
+++ b/src/diffusers/pipelines/flux/pipeline_flux_inpaint.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -28,20 +27,18 @@
 )
 
 from ...image_processor import PipelineImageInput, VaeImageProcessor
-from ...loaders import FluxIPAdapterMixin, FluxLoraLoaderMixin, TextualInversionLoaderMixin
+from ...loaders import FluxIPAdapterMixin, FluxLoraLoaderMixin
 from ...models.autoencoders import AutoencoderKL
 from ...models.transformers import FluxTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
 from ...utils import (
-    USE_PEFT_BACKEND,
     is_torch_xla_available,
     logging,
     replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
 )
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
+from .pipeline_flux_utils import FluxMixin, calculate_shift, retrieve_latents, retrieve_timesteps
 from .pipeline_output import FluxPipelineOutput
 
 
@@ -75,95 +72,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
-class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FluxIPAdapterMixin):
+class FluxInpaintPipeline(DiffusionPipeline, FluxMixin, FluxLoraLoaderMixin, FluxIPAdapterMixin):
     r"""
     The Flux pipeline for image inpainting.
 
@@ -238,181 +147,6 @@ def __init__(
         )
         self.default_sample_size = 128
 
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_t5_prompt_embeds
-    def _get_t5_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        num_images_per_prompt: int = 1,
-        max_sequence_length: int = 512,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer_2)
-
-        text_inputs = self.tokenizer_2(
-            prompt,
-            padding="max_length",
-            max_length=max_sequence_length,
-            truncation=True,
-            return_length=False,
-            return_overflowing_tokens=False,
-            return_tensors="pt",
-        )
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer_2(prompt, padding="longest", return_tensors="pt").input_ids
-
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer_2.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because `max_sequence_length` is set to "
-                f" {max_sequence_length} tokens: {removed_text}"
-            )
-
-        prompt_embeds = self.text_encoder_2(text_input_ids.to(device), output_hidden_states=False)[0]
-
-        dtype = self.text_encoder_2.dtype
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        _, seq_len, _ = prompt_embeds.shape
-
-        # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_clip_prompt_embeds
-    def _get_clip_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]],
-        num_images_per_prompt: int = 1,
-        device: Optional[torch.device] = None,
-    ):
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-        text_inputs = self.tokenizer(
-            prompt,
-            padding="max_length",
-            max_length=self.tokenizer_max_length,
-            truncation=True,
-            return_overflowing_tokens=False,
-            return_length=False,
-            return_tensors="pt",
-        )
-
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because CLIP can only handle sequences up to"
-                f" {self.tokenizer_max_length} tokens: {removed_text}"
-            )
-        prompt_embeds = self.text_encoder(text_input_ids.to(device), output_hidden_states=False)
-
-        # Use pooled output of CLIPTextModel
-        prompt_embeds = prompt_embeds.pooler_output
-        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
-
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        prompt_2: Optional[Union[str, List[str]]] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
-        max_sequence_length: int = 512,
-        lora_scale: Optional[float] = None,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                used in all text-encoders
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-        """
-        device = device or self._execution_device
-
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, FluxLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder, lora_scale)
-            if self.text_encoder_2 is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder_2, lora_scale)
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        if prompt_embeds is None:
-            prompt_2 = prompt_2 or prompt
-            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
-
-            # We only use the pooled prompt output from the CLIPTextModel
-            pooled_prompt_embeds = self._get_clip_prompt_embeds(
-                prompt=prompt,
-                device=device,
-                num_images_per_prompt=num_images_per_prompt,
-            )
-            prompt_embeds = self._get_t5_prompt_embeds(
-                prompt=prompt_2,
-                num_images_per_prompt=num_images_per_prompt,
-                max_sequence_length=max_sequence_length,
-                device=device,
-            )
-
-        if self.text_encoder is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        if self.text_encoder_2 is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder_2, lora_scale)
-
-        dtype = self.text_encoder.dtype if self.text_encoder is not None else self.transformer.dtype
-        text_ids = torch.zeros(prompt_embeds.shape[1], 3).to(device=device, dtype=dtype)
-
-        return prompt_embeds, pooled_prompt_embeds, text_ids
-
     # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_image
     def encode_image(self, image, device, num_images_per_prompt):
         dtype = next(self.image_encoder.parameters()).dtype
@@ -587,47 +321,6 @@ def check_inputs(
         if max_sequence_length is not None and max_sequence_length > 512:
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
-    def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
-        latent_image_ids = torch.zeros(height, width, 3)
-        latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
-        latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width)[None, :]
-
-        latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
-
-        latent_image_ids = latent_image_ids.reshape(
-            latent_image_id_height * latent_image_id_width, latent_image_id_channels
-        )
-
-        return latent_image_ids.to(device=device, dtype=dtype)
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._unpack_latents
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), height, width)
-
-        return latents
-
     def prepare_latents(
         self,
         image,
@@ -756,22 +449,6 @@ def prepare_mask_latents(
 
         return mask, masked_image_latents
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def joint_attention_kwargs(self):
-        return self._joint_attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/flux/pipeline_flux_kontext.py b/src/diffusers/pipelines/flux/pipeline_flux_kontext.py
index 94ae460afcd0..29ddfa333db3 100644
--- a/src/diffusers/pipelines/flux/pipeline_flux_kontext.py
+++ b/src/diffusers/pipelines/flux/pipeline_flux_kontext.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -30,17 +29,10 @@
 from ...loaders import FluxIPAdapterMixin, FluxLoraLoaderMixin, FromSingleFileMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, FluxTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import (
-    USE_PEFT_BACKEND,
-    deprecate,
-    is_torch_xla_available,
-    logging,
-    replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
-)
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
+from .pipeline_flux_utils import FluxMixin, calculate_shift, retrieve_latents, retrieve_timesteps
 from .pipeline_output import FluxPipelineOutput
 
 
@@ -101,95 +93,9 @@
 ]
 
 
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
 class FluxKontextPipeline(
     DiffusionPipeline,
+    FluxMixin,
     FluxLoraLoaderMixin,
     FromSingleFileMixin,
     TextualInversionLoaderMixin,
@@ -260,181 +166,6 @@ def __init__(
         )
         self.default_sample_size = 128
 
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_t5_prompt_embeds
-    def _get_t5_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        num_images_per_prompt: int = 1,
-        max_sequence_length: int = 512,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer_2)
-
-        text_inputs = self.tokenizer_2(
-            prompt,
-            padding="max_length",
-            max_length=max_sequence_length,
-            truncation=True,
-            return_length=False,
-            return_overflowing_tokens=False,
-            return_tensors="pt",
-        )
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer_2(prompt, padding="longest", return_tensors="pt").input_ids
-
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer_2.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because `max_sequence_length` is set to "
-                f" {max_sequence_length} tokens: {removed_text}"
-            )
-
-        prompt_embeds = self.text_encoder_2(text_input_ids.to(device), output_hidden_states=False)[0]
-
-        dtype = self.text_encoder_2.dtype
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        _, seq_len, _ = prompt_embeds.shape
-
-        # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_clip_prompt_embeds
-    def _get_clip_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]],
-        num_images_per_prompt: int = 1,
-        device: Optional[torch.device] = None,
-    ):
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-        text_inputs = self.tokenizer(
-            prompt,
-            padding="max_length",
-            max_length=self.tokenizer_max_length,
-            truncation=True,
-            return_overflowing_tokens=False,
-            return_length=False,
-            return_tensors="pt",
-        )
-
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because CLIP can only handle sequences up to"
-                f" {self.tokenizer_max_length} tokens: {removed_text}"
-            )
-        prompt_embeds = self.text_encoder(text_input_ids.to(device), output_hidden_states=False)
-
-        # Use pooled output of CLIPTextModel
-        prompt_embeds = prompt_embeds.pooler_output
-        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
-
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        prompt_2: Optional[Union[str, List[str]]] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
-        max_sequence_length: int = 512,
-        lora_scale: Optional[float] = None,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                used in all text-encoders
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-        """
-        device = device or self._execution_device
-
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, FluxLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder, lora_scale)
-            if self.text_encoder_2 is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder_2, lora_scale)
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        if prompt_embeds is None:
-            prompt_2 = prompt_2 or prompt
-            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
-
-            # We only use the pooled prompt output from the CLIPTextModel
-            pooled_prompt_embeds = self._get_clip_prompt_embeds(
-                prompt=prompt,
-                device=device,
-                num_images_per_prompt=num_images_per_prompt,
-            )
-            prompt_embeds = self._get_t5_prompt_embeds(
-                prompt=prompt_2,
-                num_images_per_prompt=num_images_per_prompt,
-                max_sequence_length=max_sequence_length,
-                device=device,
-            )
-
-        if self.text_encoder is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        if self.text_encoder_2 is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder_2, lora_scale)
-
-        dtype = self.text_encoder.dtype if self.text_encoder is not None else self.transformer.dtype
-        text_ids = torch.zeros(prompt_embeds.shape[1], 3).to(device=device, dtype=dtype)
-
-        return prompt_embeds, pooled_prompt_embeds, text_ids
-
     # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_image
     def encode_image(self, image, device, num_images_per_prompt):
         dtype = next(self.image_encoder.parameters()).dtype
@@ -554,47 +285,6 @@ def check_inputs(
         if max_sequence_length is not None and max_sequence_length > 512:
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
-    def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
-        latent_image_ids = torch.zeros(height, width, 3)
-        latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
-        latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width)[None, :]
-
-        latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
-
-        latent_image_ids = latent_image_ids.reshape(
-            latent_image_id_height * latent_image_id_width, latent_image_id_channels
-        )
-
-        return latent_image_ids.to(device=device, dtype=dtype)
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._unpack_latents
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), height, width)
-
-        return latents
-
     def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
         if isinstance(generator, list):
             image_latents = [
@@ -609,63 +299,6 @@ def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
 
         return image_latents
 
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.enable_vae_slicing
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
-        deprecate(
-            "enable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_slicing()
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.disable_vae_slicing
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
-        deprecate(
-            "disable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_slicing()
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.enable_vae_tiling
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
-        deprecate(
-            "enable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_tiling()
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.disable_vae_tiling
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
-        deprecate(
-            "disable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_tiling()
-
     def prepare_latents(
         self,
         image: Optional[torch.Tensor],
@@ -728,22 +361,6 @@ def prepare_latents(
 
         return latents, image_latents, latent_ids, image_ids
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def joint_attention_kwargs(self):
-        return self._joint_attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def current_timestep(self):
-        return self._current_timestep
-
     @property
     def interrupt(self):
         return self._interrupt
diff --git a/src/diffusers/pipelines/flux/pipeline_flux_kontext_inpaint.py b/src/diffusers/pipelines/flux/pipeline_flux_kontext_inpaint.py
index b6f957981e14..b421afeb0d52 100644
--- a/src/diffusers/pipelines/flux/pipeline_flux_kontext_inpaint.py
+++ b/src/diffusers/pipelines/flux/pipeline_flux_kontext_inpaint.py
@@ -1,7 +1,6 @@
 # Copyright 2025 ZenAI. All rights reserved.
 # author: @vuongminh1907
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -20,17 +19,10 @@
 from ...loaders import FluxIPAdapterMixin, FluxLoraLoaderMixin, FromSingleFileMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, FluxTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import (
-    USE_PEFT_BACKEND,
-    deprecate,
-    is_torch_xla_available,
-    logging,
-    replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
-)
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
+from .pipeline_flux_utils import FluxMixin, calculate_shift, retrieve_latents, retrieve_timesteps
 from .pipeline_output import FluxPipelineOutput
 
 
@@ -125,95 +117,9 @@
 ]
 
 
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
 class FluxKontextInpaintPipeline(
     DiffusionPipeline,
+    FluxMixin,
     FluxLoraLoaderMixin,
     FromSingleFileMixin,
     TextualInversionLoaderMixin,
@@ -293,181 +199,6 @@ def __init__(
         )
         self.default_sample_size = 128
 
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_t5_prompt_embeds
-    def _get_t5_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        num_images_per_prompt: int = 1,
-        max_sequence_length: int = 512,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer_2)
-
-        text_inputs = self.tokenizer_2(
-            prompt,
-            padding="max_length",
-            max_length=max_sequence_length,
-            truncation=True,
-            return_length=False,
-            return_overflowing_tokens=False,
-            return_tensors="pt",
-        )
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer_2(prompt, padding="longest", return_tensors="pt").input_ids
-
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer_2.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because `max_sequence_length` is set to "
-                f" {max_sequence_length} tokens: {removed_text}"
-            )
-
-        prompt_embeds = self.text_encoder_2(text_input_ids.to(device), output_hidden_states=False)[0]
-
-        dtype = self.text_encoder_2.dtype
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        _, seq_len, _ = prompt_embeds.shape
-
-        # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_clip_prompt_embeds
-    def _get_clip_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]],
-        num_images_per_prompt: int = 1,
-        device: Optional[torch.device] = None,
-    ):
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-        text_inputs = self.tokenizer(
-            prompt,
-            padding="max_length",
-            max_length=self.tokenizer_max_length,
-            truncation=True,
-            return_overflowing_tokens=False,
-            return_length=False,
-            return_tensors="pt",
-        )
-
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because CLIP can only handle sequences up to"
-                f" {self.tokenizer_max_length} tokens: {removed_text}"
-            )
-        prompt_embeds = self.text_encoder(text_input_ids.to(device), output_hidden_states=False)
-
-        # Use pooled output of CLIPTextModel
-        prompt_embeds = prompt_embeds.pooler_output
-        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
-
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        prompt_2: Optional[Union[str, List[str]]] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
-        max_sequence_length: int = 512,
-        lora_scale: Optional[float] = None,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                used in all text-encoders
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-        """
-        device = device or self._execution_device
-
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, FluxLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder, lora_scale)
-            if self.text_encoder_2 is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder_2, lora_scale)
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        if prompt_embeds is None:
-            prompt_2 = prompt_2 or prompt
-            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
-
-            # We only use the pooled prompt output from the CLIPTextModel
-            pooled_prompt_embeds = self._get_clip_prompt_embeds(
-                prompt=prompt,
-                device=device,
-                num_images_per_prompt=num_images_per_prompt,
-            )
-            prompt_embeds = self._get_t5_prompt_embeds(
-                prompt=prompt_2,
-                num_images_per_prompt=num_images_per_prompt,
-                max_sequence_length=max_sequence_length,
-                device=device,
-            )
-
-        if self.text_encoder is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        if self.text_encoder_2 is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder_2, lora_scale)
-
-        dtype = self.text_encoder.dtype if self.text_encoder is not None else self.transformer.dtype
-        text_ids = torch.zeros(prompt_embeds.shape[1], 3).to(device=device, dtype=dtype)
-
-        return prompt_embeds, pooled_prompt_embeds, text_ids
-
     # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_image
     def encode_image(self, image, device, num_images_per_prompt):
         dtype = next(self.image_encoder.parameters()).dtype
@@ -628,47 +359,6 @@ def check_inputs(
         if max_sequence_length is not None and max_sequence_length > 512:
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
-    def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
-        latent_image_ids = torch.zeros(height, width, 3)
-        latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
-        latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width)[None, :]
-
-        latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
-
-        latent_image_ids = latent_image_ids.reshape(
-            latent_image_id_height * latent_image_id_width, latent_image_id_channels
-        )
-
-        return latent_image_ids.to(device=device, dtype=dtype)
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._unpack_latents
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), height, width)
-
-        return latents
-
     def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
         if isinstance(generator, list):
             image_latents = [
@@ -683,63 +373,6 @@ def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
 
         return image_latents
 
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.enable_vae_slicing
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
-        deprecate(
-            "enable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_slicing()
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.disable_vae_slicing
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
-        deprecate(
-            "disable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_slicing()
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.enable_vae_tiling
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
-        deprecate(
-            "enable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_tiling()
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.disable_vae_tiling
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
-        deprecate(
-            "disable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_tiling()
-
     def prepare_latents(
         self,
         image: Optional[torch.Tensor],
@@ -917,22 +550,6 @@ def prepare_mask_latents(
 
         return mask, masked_image_latents
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def joint_attention_kwargs(self):
-        return self._joint_attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def current_timestep(self):
-        return self._current_timestep
-
     @property
     def interrupt(self):
         return self._interrupt
diff --git a/src/diffusers/pipelines/flux/pipeline_flux_prior_redux.py b/src/diffusers/pipelines/flux/pipeline_flux_prior_redux.py
index e79db337b2e3..5a00edb6b13d 100644
--- a/src/diffusers/pipelines/flux/pipeline_flux_prior_redux.py
+++ b/src/diffusers/pipelines/flux/pipeline_flux_prior_redux.py
@@ -27,17 +27,14 @@
 )
 
 from ...image_processor import PipelineImageInput
-from ...loaders import FluxLoraLoaderMixin, TextualInversionLoaderMixin
 from ...utils import (
-    USE_PEFT_BACKEND,
     is_torch_xla_available,
     logging,
     replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
 )
 from ..pipeline_utils import DiffusionPipeline
 from .modeling_flux import ReduxImageEncoder
+from .pipeline_flux_utils import FluxMixin
 from .pipeline_output import FluxPriorReduxPipelineOutput
 
 
@@ -81,7 +78,7 @@
 """
 
 
-class FluxPriorReduxPipeline(DiffusionPipeline):
+class FluxPriorReduxPipeline(DiffusionPipeline, FluxMixin):
     r"""
     The Flux Redux pipeline for image-to-image generation.
 
@@ -193,181 +190,6 @@ def encode_image(self, image, device, num_images_per_prompt):
 
         return image_enc_hidden_states
 
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_t5_prompt_embeds
-    def _get_t5_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        num_images_per_prompt: int = 1,
-        max_sequence_length: int = 512,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer_2)
-
-        text_inputs = self.tokenizer_2(
-            prompt,
-            padding="max_length",
-            max_length=max_sequence_length,
-            truncation=True,
-            return_length=False,
-            return_overflowing_tokens=False,
-            return_tensors="pt",
-        )
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer_2(prompt, padding="longest", return_tensors="pt").input_ids
-
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer_2.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because `max_sequence_length` is set to "
-                f" {max_sequence_length} tokens: {removed_text}"
-            )
-
-        prompt_embeds = self.text_encoder_2(text_input_ids.to(device), output_hidden_states=False)[0]
-
-        dtype = self.text_encoder_2.dtype
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        _, seq_len, _ = prompt_embeds.shape
-
-        # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_clip_prompt_embeds
-    def _get_clip_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]],
-        num_images_per_prompt: int = 1,
-        device: Optional[torch.device] = None,
-    ):
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt)
-
-        if isinstance(self, TextualInversionLoaderMixin):
-            prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-        text_inputs = self.tokenizer(
-            prompt,
-            padding="max_length",
-            max_length=self.tokenizer_max_length,
-            truncation=True,
-            return_overflowing_tokens=False,
-            return_length=False,
-            return_tensors="pt",
-        )
-
-        text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
-            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
-            logger.warning(
-                "The following part of your input was truncated because CLIP can only handle sequences up to"
-                f" {self.tokenizer_max_length} tokens: {removed_text}"
-            )
-        prompt_embeds = self.text_encoder(text_input_ids.to(device), output_hidden_states=False)
-
-        # Use pooled output of CLIPTextModel
-        prompt_embeds = prompt_embeds.pooler_output
-        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
-
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        prompt_2: Optional[Union[str, List[str]]] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
-        max_sequence_length: int = 512,
-        lora_scale: Optional[float] = None,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                used in all text-encoders
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-        """
-        device = device or self._execution_device
-
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, FluxLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder, lora_scale)
-            if self.text_encoder_2 is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder_2, lora_scale)
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        if prompt_embeds is None:
-            prompt_2 = prompt_2 or prompt
-            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
-
-            # We only use the pooled prompt output from the CLIPTextModel
-            pooled_prompt_embeds = self._get_clip_prompt_embeds(
-                prompt=prompt,
-                device=device,
-                num_images_per_prompt=num_images_per_prompt,
-            )
-            prompt_embeds = self._get_t5_prompt_embeds(
-                prompt=prompt_2,
-                num_images_per_prompt=num_images_per_prompt,
-                max_sequence_length=max_sequence_length,
-                device=device,
-            )
-
-        if self.text_encoder is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        if self.text_encoder_2 is not None:
-            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder_2, lora_scale)
-
-        dtype = self.text_encoder.dtype if self.text_encoder is not None else self.transformer.dtype
-        text_ids = torch.zeros(prompt_embeds.shape[1], 3).to(device=device, dtype=dtype)
-
-        return prompt_embeds, pooled_prompt_embeds, text_ids
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/flux/pipeline_flux_utils.py b/src/diffusers/pipelines/flux/pipeline_flux_utils.py
new file mode 100644
index 000000000000..b4cbf58a22c4
--- /dev/null
+++ b/src/diffusers/pipelines/flux/pipeline_flux_utils.py
@@ -0,0 +1,433 @@
+# Copyright 2025 Black Forest Labs and The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import inspect
+from typing import List, Optional, Union
+
+import torch
+
+from ...loaders import FluxLoraLoaderMixin, TextualInversionLoaderMixin
+from ...utils import USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers, unscale_lora_layers
+
+
+logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
+
+
+def calculate_shift(
+    image_seq_len,
+    base_seq_len: int = 256,
+    max_seq_len: int = 4096,
+    base_shift: float = 0.5,
+    max_shift: float = 1.15,
+):
+    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
+    b = base_shift - m * base_seq_len
+    mu = image_seq_len * m + b
+    return mu
+
+
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
+def retrieve_latents(
+    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
+):
+    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
+        return encoder_output.latent_dist.sample(generator)
+    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
+        return encoder_output.latent_dist.mode()
+    elif hasattr(encoder_output, "latents"):
+        return encoder_output.latents
+    else:
+        raise AttributeError("Could not access latents of provided encoder_output")
+
+
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
+def retrieve_timesteps(
+    scheduler,
+    num_inference_steps: Optional[int] = None,
+    device: Optional[Union[str, torch.device]] = None,
+    timesteps: Optional[List[int]] = None,
+    sigmas: Optional[List[float]] = None,
+    **kwargs,
+):
+    r"""
+    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
+    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
+
+    Args:
+        scheduler (`SchedulerMixin`):
+            The scheduler to get timesteps from.
+        num_inference_steps (`int`):
+            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
+            must be `None`.
+        device (`str` or `torch.device`, *optional*):
+            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
+        timesteps (`List[int]`, *optional*):
+            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
+            `num_inference_steps` and `sigmas` must be `None`.
+        sigmas (`List[float]`, *optional*):
+            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
+            `num_inference_steps` and `timesteps` must be `None`.
+
+    Returns:
+        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
+        second element is the number of inference steps.
+    """
+    if timesteps is not None and sigmas is not None:
+        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
+    if timesteps is not None:
+        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accepts_timesteps:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" timestep schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    elif sigmas is not None:
+        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accept_sigmas:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" sigmas schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    else:
+        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+    return timesteps, num_inference_steps
+
+
+class FluxMixin:
+    @property
+    def guidance_scale(self):
+        return self._guidance_scale
+
+    @property
+    def joint_attention_kwargs(self):
+        return self._joint_attention_kwargs
+
+    @property
+    def num_timesteps(self):
+        return self._num_timesteps
+
+    @property
+    def current_timestep(self):
+        return self._current_timestep
+
+    @property
+    def interrupt(self):
+        return self._interrupt
+
+    def enable_vae_slicing(self):
+        r"""
+        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
+        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
+        """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
+        self.vae.enable_slicing()
+
+    def disable_vae_slicing(self):
+        r"""
+        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
+        computing decoding in one step.
+        """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
+        self.vae.disable_slicing()
+
+    def enable_vae_tiling(self):
+        r"""
+        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
+        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
+        processing larger images.
+        """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
+        self.vae.enable_tiling()
+
+    def disable_vae_tiling(self):
+        r"""
+        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
+        computing decoding in one step.
+        """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
+        self.vae.disable_tiling()
+
+    def encode_prompt(
+        self,
+        prompt: Union[str, List[str]],
+        prompt_2: Optional[Union[str, List[str]]] = None,
+        device: Optional[torch.device] = None,
+        num_images_per_prompt: int = 1,
+        prompt_embeds: Optional[torch.FloatTensor] = None,
+        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
+        max_sequence_length: int = 512,
+        lora_scale: Optional[float] = None,
+    ):
+        r"""
+
+        Args:
+            prompt (`str` or `List[str]`, *optional*):
+                prompt to be encoded
+            prompt_2 (`str` or `List[str]`, *optional*):
+                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
+                used in all text-encoders
+            device: (`torch.device`):
+                torch device
+            num_images_per_prompt (`int`):
+                number of images that should be generated per prompt
+            prompt_embeds (`torch.FloatTensor`, *optional*):
+                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
+                provided, text embeddings will be generated from `prompt` input argument.
+            pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
+                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
+                If not provided, pooled text embeddings will be generated from `prompt` input argument.
+            lora_scale (`float`, *optional*):
+                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
+        """
+        device = device or self._execution_device
+
+        # set lora scale so that monkey patched LoRA
+        # function of text encoder can correctly access it
+        if lora_scale is not None and isinstance(self, FluxLoraLoaderMixin):
+            self._lora_scale = lora_scale
+
+            # dynamically adjust the LoRA scale
+            if self.text_encoder is not None and USE_PEFT_BACKEND:
+                scale_lora_layers(self.text_encoder, lora_scale)
+            if self.text_encoder_2 is not None and USE_PEFT_BACKEND:
+                scale_lora_layers(self.text_encoder_2, lora_scale)
+
+        prompt = [prompt] if isinstance(prompt, str) else prompt
+
+        if prompt_embeds is None:
+            prompt_2 = prompt_2 or prompt
+            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
+
+            # We only use the pooled prompt output from the CLIPTextModel
+            pooled_prompt_embeds = self._get_clip_prompt_embeds(
+                prompt=prompt,
+                device=device,
+                num_images_per_prompt=num_images_per_prompt,
+            )
+            prompt_embeds = self._get_t5_prompt_embeds(
+                prompt=prompt_2,
+                num_images_per_prompt=num_images_per_prompt,
+                max_sequence_length=max_sequence_length,
+                device=device,
+            )
+
+        if self.text_encoder is not None:
+            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
+                # Retrieve the original scale by scaling back the LoRA layers
+                unscale_lora_layers(self.text_encoder, lora_scale)
+
+        if self.text_encoder_2 is not None:
+            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
+                # Retrieve the original scale by scaling back the LoRA layers
+                unscale_lora_layers(self.text_encoder_2, lora_scale)
+
+        dtype = self.text_encoder.dtype if self.text_encoder is not None else self.transformer.dtype
+        text_ids = torch.zeros(prompt_embeds.shape[1], 3).to(device=device, dtype=dtype)
+
+        return prompt_embeds, pooled_prompt_embeds, text_ids
+
+    @staticmethod
+    def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
+        latent_image_ids = torch.zeros(height, width, 3)
+        latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
+        latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width)[None, :]
+
+        latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
+
+        latent_image_ids = latent_image_ids.reshape(
+            latent_image_id_height * latent_image_id_width, latent_image_id_channels
+        )
+
+        return latent_image_ids.to(device=device, dtype=dtype)
+
+    @staticmethod
+    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
+        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
+        latents = latents.permute(0, 2, 4, 1, 3, 5)
+        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
+
+        return latents
+
+    @staticmethod
+    def _unpack_latents(latents, height, width, vae_scale_factor):
+        batch_size, num_patches, channels = latents.shape
+
+        # VAE applies 8x compression on images but we must also account for packing which requires
+        # latent height and width to be divisible by 2.
+        height = 2 * (int(height) // (vae_scale_factor * 2))
+        width = 2 * (int(width) // (vae_scale_factor * 2))
+
+        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
+        latents = latents.permute(0, 3, 1, 4, 2, 5)
+
+        latents = latents.reshape(batch_size, channels // (2 * 2), height, width)
+
+        return latents
+
+    def _get_t5_prompt_embeds(
+        self,
+        prompt: Union[str, List[str]] = None,
+        num_images_per_prompt: int = 1,
+        max_sequence_length: int = 512,
+        device: Optional[torch.device] = None,
+        dtype: Optional[torch.dtype] = None,
+    ):
+        device = device or self._execution_device
+        dtype = dtype or self.text_encoder.dtype
+
+        prompt = [prompt] if isinstance(prompt, str) else prompt
+        batch_size = len(prompt)
+
+        if isinstance(self, TextualInversionLoaderMixin):
+            prompt = self.maybe_convert_prompt(prompt, self.tokenizer_2)
+
+        text_inputs = self.tokenizer_2(
+            prompt,
+            padding="max_length",
+            max_length=max_sequence_length,
+            truncation=True,
+            return_length=False,
+            return_overflowing_tokens=False,
+            return_tensors="pt",
+        )
+        text_input_ids = text_inputs.input_ids
+        untruncated_ids = self.tokenizer_2(prompt, padding="longest", return_tensors="pt").input_ids
+
+        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
+            removed_text = self.tokenizer_2.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
+            logger.warning(
+                "The following part of your input was truncated because `max_sequence_length` is set to "
+                f" {max_sequence_length} tokens: {removed_text}"
+            )
+
+        prompt_embeds = self.text_encoder_2(text_input_ids.to(device), output_hidden_states=False)[0]
+
+        dtype = self.text_encoder_2.dtype
+        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
+
+        _, seq_len, _ = prompt_embeds.shape
+
+        # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
+        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
+        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
+
+        return prompt_embeds
+
+    def _get_clip_prompt_embeds(
+        self,
+        prompt: Union[str, List[str]],
+        num_images_per_prompt: int = 1,
+        device: Optional[torch.device] = None,
+    ):
+        device = device or self._execution_device
+
+        prompt = [prompt] if isinstance(prompt, str) else prompt
+        batch_size = len(prompt)
+
+        if isinstance(self, TextualInversionLoaderMixin):
+            prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
+
+        text_inputs = self.tokenizer(
+            prompt,
+            padding="max_length",
+            max_length=self.tokenizer_max_length,
+            truncation=True,
+            return_overflowing_tokens=False,
+            return_length=False,
+            return_tensors="pt",
+        )
+
+        text_input_ids = text_inputs.input_ids
+        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
+        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
+            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1 : -1])
+            logger.warning(
+                "The following part of your input was truncated because CLIP can only handle sequences up to"
+                f" {self.tokenizer_max_length} tokens: {removed_text}"
+            )
+        prompt_embeds = self.text_encoder(text_input_ids.to(device), output_hidden_states=False)
+
+        # Use pooled output of CLIPTextModel
+        prompt_embeds = prompt_embeds.pooler_output
+        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
+
+        # duplicate text embeddings for each generation per prompt, using mps friendly method
+        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt)
+        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)
+
+        return prompt_embeds
+
+
+class FluxControlMixin(FluxMixin):
+    # Copied from diffusers.pipelines.controlnet_sd3.pipeline_stable_diffusion_3_controlnet.StableDiffusion3ControlNetPipeline.prepare_image
+    def prepare_image(
+        self,
+        image,
+        width,
+        height,
+        batch_size,
+        num_images_per_prompt,
+        device,
+        dtype,
+        do_classifier_free_guidance=False,
+        guess_mode=False,
+    ):
+        if isinstance(image, torch.Tensor):
+            pass
+        else:
+            image = self.image_processor.preprocess(image, height=height, width=width)
+
+        image_batch_size = image.shape[0]
+
+        if image_batch_size == 1:
+            repeat_by = batch_size
+        else:
+            # image batch size is the same as prompt batch size
+            repeat_by = num_images_per_prompt
+
+        image = image.repeat_interleave(repeat_by, dim=0)
+
+        image = image.to(device=device, dtype=dtype)
+
+        if do_classifier_free_guidance and not guess_mode:
+            image = torch.cat([image] * 2)
+
+        return image
diff --git a/src/diffusers/pipelines/flux2/pipeline_flux2.py b/src/diffusers/pipelines/flux2/pipeline_flux2.py
index 72d6ef739aae..7822ba93034c 100644
--- a/src/diffusers/pipelines/flux2/pipeline_flux2.py
+++ b/src/diffusers/pipelines/flux2/pipeline_flux2.py
@@ -93,7 +93,7 @@ def compute_empirical_mu(image_seq_len: int, num_steps: int) -> float:
     return float(mu)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -153,7 +153,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/pipelines/hidream_image/pipeline_hidream_image.py b/src/diffusers/pipelines/hidream_image/pipeline_hidream_image.py
index b6af23bca8fd..336a143e96bc 100644
--- a/src/diffusers/pipelines/hidream_image/pipeline_hidream_image.py
+++ b/src/diffusers/pipelines/hidream_image/pipeline_hidream_image.py
@@ -83,7 +83,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -97,7 +97,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/hunyuan_image/pipeline_hunyuanimage.py b/src/diffusers/pipelines/hunyuan_image/pipeline_hunyuanimage.py
index 658935ccd886..28dac10fd6e7 100644
--- a/src/diffusers/pipelines/hunyuan_image/pipeline_hunyuanimage.py
+++ b/src/diffusers/pipelines/hunyuan_image/pipeline_hunyuanimage.py
@@ -95,7 +95,7 @@ def extract_glyph_text(prompt: str):
     return glyph_text_formatted
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/hunyuan_image/pipeline_hunyuanimage_refiner.py b/src/diffusers/pipelines/hunyuan_image/pipeline_hunyuanimage_refiner.py
index f38f53d9a562..27f251ad8e17 100644
--- a/src/diffusers/pipelines/hunyuan_image/pipeline_hunyuanimage_refiner.py
+++ b/src/diffusers/pipelines/hunyuan_image/pipeline_hunyuanimage_refiner.py
@@ -59,7 +59,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -119,7 +119,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_skyreels_image2video.py b/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_skyreels_image2video.py
index b50a6ae3ed27..8a0b3cf86547 100644
--- a/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_skyreels_image2video.py
+++ b/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_skyreels_image2video.py
@@ -92,7 +92,7 @@
 }
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -152,7 +152,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video.py b/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video.py
index 5c8e295eaf4c..42127238aa1c 100644
--- a/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video.py
+++ b/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video.py
@@ -81,7 +81,7 @@
 }
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video_framepack.py b/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video_framepack.py
index 8006514f47ea..e62c8bbf880a 100644
--- a/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video_framepack.py
+++ b/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video_framepack.py
@@ -161,7 +161,7 @@
 }
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -175,7 +175,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video_image2video.py b/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video_image2video.py
index aa04e6509730..11d33b871e94 100644
--- a/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video_image2video.py
+++ b/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video_image2video.py
@@ -144,7 +144,7 @@ def _expand_input_ids_with_image_tokens(
     }
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -204,7 +204,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/pipelines/hunyuandit/pipeline_hunyuandit.py b/src/diffusers/pipelines/hunyuandit/pipeline_hunyuandit.py
index e2f935aaf4b9..9b2821f1bd8f 100644
--- a/src/diffusers/pipelines/hunyuandit/pipeline_hunyuandit.py
+++ b/src/diffusers/pipelines/hunyuandit/pipeline_hunyuandit.py
@@ -119,7 +119,7 @@ def get_resize_crop_region_for_grid(src, tgt_size):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -410,7 +410,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, prompt_attention_mask, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -425,7 +425,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/i2vgen_xl/pipeline_i2vgen_xl.py b/src/diffusers/pipelines/i2vgen_xl/pipeline_i2vgen_xl.py
index c6cc724a71f0..d2da8ec7b246 100644
--- a/src/diffusers/pipelines/i2vgen_xl/pipeline_i2vgen_xl.py
+++ b/src/diffusers/pipelines/i2vgen_xl/pipeline_i2vgen_xl.py
@@ -382,7 +382,7 @@ def decode_latents(self, latents, decode_chunk_size=None):
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/kandinsky3/pipeline_kandinsky3_img2img.py b/src/diffusers/pipelines/kandinsky3/pipeline_kandinsky3_img2img.py
index 73c268897502..519ca5555d16 100644
--- a/src/diffusers/pipelines/kandinsky3/pipeline_kandinsky3_img2img.py
+++ b/src/diffusers/pipelines/kandinsky3/pipeline_kandinsky3_img2img.py
@@ -295,7 +295,7 @@ def prepare_latents(self, image, timestep, batch_size, num_images_per_prompt, dt
 
         return latents
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/kolors/pipeline_kolors.py b/src/diffusers/pipelines/kolors/pipeline_kolors.py
index 948f73ed91eb..cb6d7cbdaacf 100644
--- a/src/diffusers/pipelines/kolors/pipeline_kolors.py
+++ b/src/diffusers/pipelines/kolors/pipeline_kolors.py
@@ -61,7 +61,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -432,7 +432,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -578,17 +578,13 @@ def _get_add_time_ids(
         add_time_ids = torch.tensor([add_time_ids], dtype=dtype)
         return add_time_ids
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.upcast_vae
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.upcast_vae
     def upcast_vae(self):
         dtype = self.vae.dtype
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-                FusedAttnProcessor2_0,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor, FusedAttnProcessor2_0),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/pipelines/kolors/pipeline_kolors_img2img.py b/src/diffusers/pipelines/kolors/pipeline_kolors_img2img.py
index 67d49b9a8c5e..b43c874c092b 100644
--- a/src/diffusers/pipelines/kolors/pipeline_kolors_img2img.py
+++ b/src/diffusers/pipelines/kolors/pipeline_kolors_img2img.py
@@ -66,7 +66,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -80,7 +80,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -452,7 +452,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -706,17 +706,13 @@ def _get_add_time_ids(
         add_time_ids = torch.tensor([add_time_ids], dtype=dtype)
         return add_time_ids
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.upcast_vae
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.upcast_vae
     def upcast_vae(self):
         dtype = self.vae.dtype
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-                FusedAttnProcessor2_0,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor, FusedAttnProcessor2_0),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/pipelines/latent_consistency_models/pipeline_latent_consistency_img2img.py b/src/diffusers/pipelines/latent_consistency_models/pipeline_latent_consistency_img2img.py
index 59f733a498ed..35920f192054 100644
--- a/src/diffusers/pipelines/latent_consistency_models/pipeline_latent_consistency_img2img.py
+++ b/src/diffusers/pipelines/latent_consistency_models/pipeline_latent_consistency_img2img.py
@@ -51,7 +51,7 @@
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -65,7 +65,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -237,7 +237,7 @@ def __init__(
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -491,7 +491,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -603,7 +603,7 @@ def get_guidance_scale_embedding(
         assert emb.shape == (w.shape[0], embedding_dim)
         return emb
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/latent_consistency_models/pipeline_latent_consistency_text2img.py b/src/diffusers/pipelines/latent_consistency_models/pipeline_latent_consistency_text2img.py
index e463884618f5..5c8c7399fd19 100644
--- a/src/diffusers/pipelines/latent_consistency_models/pipeline_latent_consistency_text2img.py
+++ b/src/diffusers/pipelines/latent_consistency_models/pipeline_latent_consistency_text2img.py
@@ -70,7 +70,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -222,7 +222,7 @@ def __init__(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -476,7 +476,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -544,7 +544,7 @@ def get_guidance_scale_embedding(
         assert emb.shape == (w.shape[0], embedding_dim)
         return emb
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/latte/pipeline_latte.py b/src/diffusers/pipelines/latte/pipeline_latte.py
index 4d42a7049ec9..523f93c5563b 100644
--- a/src/diffusers/pipelines/latte/pipeline_latte.py
+++ b/src/diffusers/pipelines/latte/pipeline_latte.py
@@ -77,7 +77,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -352,7 +352,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion.py b/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion.py
index fbf4dc23d043..8f6b107245b2 100644
--- a/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion.py
+++ b/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion.py
@@ -239,7 +239,7 @@ def __call__(
         return hidden_states
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -408,7 +408,7 @@ def __init__(
 
         self.inversion_steps = None
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -423,7 +423,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -435,7 +435,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, eta, generator=None):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion_xl.py b/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion_xl.py
index c1f9a98f0632..1a7d5f016cd9 100644
--- a/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion_xl.py
+++ b/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion_xl.py
@@ -619,7 +619,7 @@ def encode_prompt(
             num_edit_tokens,
         )
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, eta, generator=None):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -692,10 +692,7 @@ def upcast_vae(self):
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
@@ -1712,7 +1709,7 @@ def invert(
         return LEditsPPInversionPipelineOutput(images=resized, vae_reconstruction_images=image_rec)
 
 
-# Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
diff --git a/src/diffusers/pipelines/ltx/pipeline_ltx.py b/src/diffusers/pipelines/ltx/pipeline_ltx.py
index 8ca8b4419e18..9246871e82b9 100644
--- a/src/diffusers/pipelines/ltx/pipeline_ltx.py
+++ b/src/diffusers/pipelines/ltx/pipeline_ltx.py
@@ -66,7 +66,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -80,7 +80,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -140,7 +140,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
diff --git a/src/diffusers/pipelines/ltx/pipeline_ltx_condition.py b/src/diffusers/pipelines/ltx/pipeline_ltx_condition.py
index 48a6f0837c8d..9df83114a5b0 100644
--- a/src/diffusers/pipelines/ltx/pipeline_ltx_condition.py
+++ b/src/diffusers/pipelines/ltx/pipeline_ltx_condition.py
@@ -134,7 +134,7 @@ def linear_quadratic_schedule(num_steps, threshold_noise=0.025, linear_steps=Non
     return torch.tensor(sigma_schedule[:-1])
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -148,7 +148,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -208,7 +208,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -222,7 +222,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
diff --git a/src/diffusers/pipelines/ltx/pipeline_ltx_image2video.py b/src/diffusers/pipelines/ltx/pipeline_ltx_image2video.py
index f30f8a3dc8f6..07ab385be511 100644
--- a/src/diffusers/pipelines/ltx/pipeline_ltx_image2video.py
+++ b/src/diffusers/pipelines/ltx/pipeline_ltx_image2video.py
@@ -71,7 +71,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -85,7 +85,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -145,7 +145,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -159,7 +159,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
diff --git a/src/diffusers/pipelines/ltx/pipeline_ltx_latent_upsample.py b/src/diffusers/pipelines/ltx/pipeline_ltx_latent_upsample.py
index 9acff105e56d..87f6b3388c70 100644
--- a/src/diffusers/pipelines/ltx/pipeline_ltx_latent_upsample.py
+++ b/src/diffusers/pipelines/ltx/pipeline_ltx_latent_upsample.py
@@ -29,7 +29,7 @@
 logger = get_logger(__name__)  # pylint: disable=invalid-name
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/pipelines/lucy/pipeline_lucy_edit.py b/src/diffusers/pipelines/lucy/pipeline_lucy_edit.py
index 69f69d5768a8..84fee449a6fe 100644
--- a/src/diffusers/pipelines/lucy/pipeline_lucy_edit.py
+++ b/src/diffusers/pipelines/lucy/pipeline_lucy_edit.py
@@ -117,7 +117,7 @@ def prompt_clean(text):
     return text
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/pipelines/lumina/pipeline_lumina.py b/src/diffusers/pipelines/lumina/pipeline_lumina.py
index b59c265646cd..c5e8aeb9ec73 100644
--- a/src/diffusers/pipelines/lumina/pipeline_lumina.py
+++ b/src/diffusers/pipelines/lumina/pipeline_lumina.py
@@ -73,7 +73,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -368,7 +368,7 @@ def encode_prompt(
 
         return prompt_embeds, prompt_attention_mask, negative_prompt_embeds, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/lumina2/pipeline_lumina2.py b/src/diffusers/pipelines/lumina2/pipeline_lumina2.py
index 937803edbcbc..64c517ee57a1 100644
--- a/src/diffusers/pipelines/lumina2/pipeline_lumina2.py
+++ b/src/diffusers/pipelines/lumina2/pipeline_lumina2.py
@@ -60,7 +60,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -74,7 +74,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -338,7 +338,7 @@ def encode_prompt(
 
         return prompt_embeds, prompt_attention_mask, negative_prompt_embeds, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/mochi/pipeline_mochi.py b/src/diffusers/pipelines/mochi/pipeline_mochi.py
index 5874a92c6f2f..c71fec4903cb 100644
--- a/src/diffusers/pipelines/mochi/pipeline_mochi.py
+++ b/src/diffusers/pipelines/mochi/pipeline_mochi.py
@@ -75,7 +75,7 @@ def linear_quadratic_schedule(num_steps, threshold_noise, linear_steps=None):
     return sigma_schedule
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/musicldm/pipeline_musicldm.py b/src/diffusers/pipelines/musicldm/pipeline_musicldm.py
index c909e5eb0d26..bfd8c502efcc 100644
--- a/src/diffusers/pipelines/musicldm/pipeline_musicldm.py
+++ b/src/diffusers/pipelines/musicldm/pipeline_musicldm.py
@@ -294,7 +294,7 @@ def score_waveforms(self, text, audio, num_waveforms_per_prompt, device, dtype):
         audio = torch.index_select(audio, 0, indices.reshape(-1).cpu())
         return audio
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/omnigen/pipeline_omnigen.py b/src/diffusers/pipelines/omnigen/pipeline_omnigen.py
index 090cb46aace4..8324bfe00c66 100644
--- a/src/diffusers/pipelines/omnigen/pipeline_omnigen.py
+++ b/src/diffusers/pipelines/omnigen/pipeline_omnigen.py
@@ -56,7 +56,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/pag/pipeline_pag_controlnet_sd.py b/src/diffusers/pipelines/pag/pipeline_pag_controlnet_sd.py
index 1abef014301a..eaee7dbc146e 100644
--- a/src/diffusers/pipelines/pag/pipeline_pag_controlnet_sd.py
+++ b/src/diffusers/pipelines/pag/pipeline_pag_controlnet_sd.py
@@ -105,7 +105,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -271,7 +271,7 @@ def __init__(
 
         self.set_pag_applied_layers(pag_applied_layers)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -525,7 +525,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -540,7 +540,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/pag/pipeline_pag_controlnet_sd_inpaint.py b/src/diffusers/pipelines/pag/pipeline_pag_controlnet_sd_inpaint.py
index 2781af789018..1819fabe2395 100644
--- a/src/diffusers/pipelines/pag/pipeline_pag_controlnet_sd_inpaint.py
+++ b/src/diffusers/pipelines/pag/pipeline_pag_controlnet_sd_inpaint.py
@@ -117,7 +117,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -247,7 +247,7 @@ def __init__(
         self.register_to_config(requires_safety_checker=requires_safety_checker)
         self.set_pag_applied_layers(pag_applied_layers)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -501,7 +501,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -516,7 +516,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/pag/pipeline_pag_controlnet_sd_xl.py b/src/diffusers/pipelines/pag/pipeline_pag_controlnet_sd_xl.py
index 24cbab43c966..5b0ff3080674 100644
--- a/src/diffusers/pipelines/pag/pipeline_pag_controlnet_sd_xl.py
+++ b/src/diffusers/pipelines/pag/pipeline_pag_controlnet_sd_xl.py
@@ -125,7 +125,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -305,7 +305,7 @@ def __init__(
         self.register_to_config(force_zeros_for_empty_prompt=force_zeros_for_empty_prompt)
         self.set_pag_applied_layers(pag_applied_layers)
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -615,7 +615,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -940,10 +940,7 @@ def upcast_vae(self):
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/pipelines/pag/pipeline_pag_controlnet_sd_xl_img2img.py b/src/diffusers/pipelines/pag/pipeline_pag_controlnet_sd_xl_img2img.py
index a6df1b22c8b9..8cc93e0e9716 100644
--- a/src/diffusers/pipelines/pag/pipeline_pag_controlnet_sd_xl_img2img.py
+++ b/src/diffusers/pipelines/pag/pipeline_pag_controlnet_sd_xl_img2img.py
@@ -151,7 +151,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -297,7 +297,7 @@ def __init__(
 
         self.set_pag_applied_layers(pag_applied_layers)
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -607,7 +607,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -1053,10 +1053,7 @@ def upcast_vae(self):
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/pipelines/pag/pipeline_pag_hunyuandit.py b/src/diffusers/pipelines/pag/pipeline_pag_hunyuandit.py
index d156eac8f3f7..91bf831d1189 100644
--- a/src/diffusers/pipelines/pag/pipeline_pag_hunyuandit.py
+++ b/src/diffusers/pipelines/pag/pipeline_pag_hunyuandit.py
@@ -122,7 +122,7 @@ def get_resize_crop_region_for_grid(src, tgt_size):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -420,7 +420,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, prompt_attention_mask, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -435,7 +435,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/pag/pipeline_pag_kolors.py b/src/diffusers/pipelines/pag/pipeline_pag_kolors.py
index 1368358db6ba..78bf5378c566 100644
--- a/src/diffusers/pipelines/pag/pipeline_pag_kolors.py
+++ b/src/diffusers/pipelines/pag/pipeline_pag_kolors.py
@@ -66,7 +66,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -449,7 +449,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -596,17 +596,13 @@ def _get_add_time_ids(
         add_time_ids = torch.tensor([add_time_ids], dtype=dtype)
         return add_time_ids
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.upcast_vae
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.upcast_vae
     def upcast_vae(self):
         dtype = self.vae.dtype
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-                FusedAttnProcessor2_0,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor, FusedAttnProcessor2_0),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/pipelines/pag/pipeline_pag_pixart_sigma.py b/src/diffusers/pipelines/pag/pipeline_pag_pixart_sigma.py
index 9031877b5b8d..4057201db469 100644
--- a/src/diffusers/pipelines/pag/pipeline_pag_pixart_sigma.py
+++ b/src/diffusers/pipelines/pag/pipeline_pag_pixart_sigma.py
@@ -81,7 +81,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -322,7 +322,7 @@ def encode_prompt(
 
         return prompt_embeds, prompt_attention_mask, negative_prompt_embeds, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/pag/pipeline_pag_sana.py b/src/diffusers/pipelines/pag/pipeline_pag_sana.py
index 9e91ccbe8006..51a1be293f59 100644
--- a/src/diffusers/pipelines/pag/pipeline_pag_sana.py
+++ b/src/diffusers/pipelines/pag/pipeline_pag_sana.py
@@ -85,7 +85,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -384,7 +384,7 @@ def encode_prompt(
 
         return prompt_embeds, prompt_attention_mask, negative_prompt_embeds, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/pag/pipeline_pag_sd.py b/src/diffusers/pipelines/pag/pipeline_pag_sd.py
index ea64f8be2c50..540bc1ec6e94 100644
--- a/src/diffusers/pipelines/pag/pipeline_pag_sd.py
+++ b/src/diffusers/pipelines/pag/pipeline_pag_sd.py
@@ -67,7 +67,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -94,7 +94,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -300,7 +300,7 @@ def __init__(
 
         self.set_pag_applied_layers(pag_applied_layers)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -554,7 +554,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -569,7 +569,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/pag/pipeline_pag_sd_3.py b/src/diffusers/pipelines/pag/pipeline_pag_sd_3.py
index 941b675099b9..493a8810c264 100644
--- a/src/diffusers/pipelines/pag/pipeline_pag_sd_3.py
+++ b/src/diffusers/pipelines/pag/pipeline_pag_sd_3.py
@@ -73,7 +73,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/pag/pipeline_pag_sd_3_img2img.py b/src/diffusers/pipelines/pag/pipeline_pag_sd_3_img2img.py
index f40dd52fc244..0d6f5a75d80f 100644
--- a/src/diffusers/pipelines/pag/pipeline_pag_sd_3_img2img.py
+++ b/src/diffusers/pipelines/pag/pipeline_pag_sd_3_img2img.py
@@ -75,7 +75,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -89,7 +89,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/pag/pipeline_pag_sd_animatediff.py b/src/diffusers/pipelines/pag/pipeline_pag_sd_animatediff.py
index de13be9c4d22..5b52e23e43b4 100644
--- a/src/diffusers/pipelines/pag/pipeline_pag_sd_animatediff.py
+++ b/src/diffusers/pipelines/pag/pipeline_pag_sd_animatediff.py
@@ -161,7 +161,7 @@ def __init__(
 
         self.set_pag_applied_layers(pag_applied_layers)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
     def encode_prompt(
         self,
         prompt,
@@ -434,7 +434,7 @@ def decode_latents(self, latents, decode_chunk_size: int = 16):
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/pag/pipeline_pag_sd_img2img.py b/src/diffusers/pipelines/pag/pipeline_pag_sd_img2img.py
index 8351112ce409..dd3d19ac0a3c 100644
--- a/src/diffusers/pipelines/pag/pipeline_pag_sd_img2img.py
+++ b/src/diffusers/pipelines/pag/pipeline_pag_sd_img2img.py
@@ -75,7 +75,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -89,7 +89,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -295,7 +295,7 @@ def __init__(
 
         self.set_pag_applied_layers(pag_applied_layers)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -549,7 +549,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -564,7 +564,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/pag/pipeline_pag_sd_inpaint.py b/src/diffusers/pipelines/pag/pipeline_pag_sd_inpaint.py
index 6b1b294e10f5..7349acd6ab39 100644
--- a/src/diffusers/pipelines/pag/pipeline_pag_sd_inpaint.py
+++ b/src/diffusers/pipelines/pag/pipeline_pag_sd_inpaint.py
@@ -80,7 +80,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -94,7 +94,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -121,7 +121,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -330,7 +330,7 @@ def __init__(
 
         self.set_pag_applied_layers(pag_applied_layers)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -584,7 +584,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -599,7 +599,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/pag/pipeline_pag_sd_xl.py b/src/diffusers/pipelines/pag/pipeline_pag_sd_xl.py
index 6b62ddcc7ca5..04b83fe927e0 100644
--- a/src/diffusers/pipelines/pag/pipeline_pag_sd_xl.py
+++ b/src/diffusers/pipelines/pag/pipeline_pag_sd_xl.py
@@ -32,11 +32,7 @@
     TextualInversionLoaderMixin,
 )
 from ...models import AutoencoderKL, ImageProjection, UNet2DConditionModel
-from ...models.attention_processor import (
-    AttnProcessor2_0,
-    FusedAttnProcessor2_0,
-    XFormersAttnProcessor,
-)
+from ...models.attention_processor import AttnProcessor2_0, FusedAttnProcessor2_0, XFormersAttnProcessor
 from ...models.lora import adjust_lora_scale_text_encoder
 from ...schedulers import KarrasDiffusionSchedulers
 from ...utils import (
@@ -86,7 +82,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -113,7 +109,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -293,7 +289,7 @@ def __init__(
 
         self.set_pag_applied_layers(pag_applied_layers)
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -603,7 +599,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -760,17 +756,13 @@ def _get_add_time_ids(
         add_time_ids = torch.tensor([add_time_ids], dtype=dtype)
         return add_time_ids
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.upcast_vae
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.upcast_vae
     def upcast_vae(self):
         dtype = self.vae.dtype
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-                FusedAttnProcessor2_0,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor, FusedAttnProcessor2_0),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/pipelines/pag/pipeline_pag_sd_xl_img2img.py b/src/diffusers/pipelines/pag/pipeline_pag_sd_xl_img2img.py
index b6422b23648c..310984c9cd77 100644
--- a/src/diffusers/pipelines/pag/pipeline_pag_sd_xl_img2img.py
+++ b/src/diffusers/pipelines/pag/pipeline_pag_sd_xl_img2img.py
@@ -90,7 +90,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -117,7 +117,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -131,7 +131,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -310,7 +310,7 @@ def __init__(
 
         self.set_pag_applied_layers(pag_applied_layers)
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -549,7 +549,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -914,10 +914,7 @@ def upcast_vae(self):
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/pipelines/pag/pipeline_pag_sd_xl_inpaint.py b/src/diffusers/pipelines/pag/pipeline_pag_sd_xl_inpaint.py
index 2a8f7a448d5e..0f5d8127d00d 100644
--- a/src/diffusers/pipelines/pag/pipeline_pag_sd_xl_inpaint.py
+++ b/src/diffusers/pipelines/pag/pipeline_pag_sd_xl_inpaint.py
@@ -103,7 +103,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -130,7 +130,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -144,7 +144,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -400,7 +400,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -639,7 +639,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -1005,10 +1005,7 @@ def upcast_vae(self):
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/pipelines/paint_by_example/pipeline_paint_by_example.py b/src/diffusers/pipelines/paint_by_example/pipeline_paint_by_example.py
index c09992befbcb..2482c16aabc1 100644
--- a/src/diffusers/pipelines/paint_by_example/pipeline_paint_by_example.py
+++ b/src/diffusers/pipelines/paint_by_example/pipeline_paint_by_example.py
@@ -41,7 +41,7 @@
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -217,7 +217,7 @@ def __init__(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -232,7 +232,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -250,7 +250,7 @@ def prepare_extra_step_kwargs(self, generator, eta):
             extra_step_kwargs["generator"] = generator
         return extra_step_kwargs
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
diff --git a/src/diffusers/pipelines/pia/pipeline_pia.py b/src/diffusers/pipelines/pia/pipeline_pia.py
index dfc6e83fbd7c..37f60df674e8 100644
--- a/src/diffusers/pipelines/pia/pipeline_pia.py
+++ b/src/diffusers/pipelines/pia/pipeline_pia.py
@@ -209,7 +209,7 @@ def __init__(
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.video_processor = VideoProcessor(do_resize=False, vae_scale_factor=self.vae_scale_factor)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
     def encode_prompt(
         self,
         prompt,
@@ -430,7 +430,7 @@ def decode_latents(self, latents):
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/pixart_alpha/pipeline_pixart_alpha.py b/src/diffusers/pipelines/pixart_alpha/pipeline_pixart_alpha.py
index 1d718a4852a4..e0fb404afffe 100644
--- a/src/diffusers/pipelines/pixart_alpha/pipeline_pixart_alpha.py
+++ b/src/diffusers/pipelines/pixart_alpha/pipeline_pixart_alpha.py
@@ -178,7 +178,7 @@
 }
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -435,7 +435,7 @@ def encode_prompt(
 
         return prompt_embeds, prompt_attention_mask, negative_prompt_embeds, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/pixart_alpha/pipeline_pixart_sigma.py b/src/diffusers/pipelines/pixart_alpha/pipeline_pixart_sigma.py
index bb169ac5c443..6950b336f374 100644
--- a/src/diffusers/pipelines/pixart_alpha/pipeline_pixart_sigma.py
+++ b/src/diffusers/pipelines/pixart_alpha/pipeline_pixart_sigma.py
@@ -122,7 +122,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -379,7 +379,7 @@ def encode_prompt(
 
         return prompt_embeds, prompt_attention_mask, negative_prompt_embeds, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/qwenimage/pipeline_qwen_utils.py b/src/diffusers/pipelines/qwenimage/pipeline_qwen_utils.py
new file mode 100644
index 000000000000..ccd7cdf7d378
--- /dev/null
+++ b/src/diffusers/pipelines/qwenimage/pipeline_qwen_utils.py
@@ -0,0 +1,441 @@
+# Copyright 2025 Qwen-Image Team and The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import inspect
+import math
+from typing import List, Optional, Union
+
+import torch
+
+from ...utils import deprecate
+
+
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
+def calculate_shift(
+    image_seq_len,
+    base_seq_len: int = 256,
+    max_seq_len: int = 4096,
+    base_shift: float = 0.5,
+    max_shift: float = 1.15,
+):
+    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
+    b = base_shift - m * base_seq_len
+    mu = image_seq_len * m + b
+    return mu
+
+
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
+def retrieve_timesteps(
+    scheduler,
+    num_inference_steps: Optional[int] = None,
+    device: Optional[Union[str, torch.device]] = None,
+    timesteps: Optional[List[int]] = None,
+    sigmas: Optional[List[float]] = None,
+    **kwargs,
+):
+    r"""
+    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
+    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
+
+    Args:
+        scheduler (`SchedulerMixin`):
+            The scheduler to get timesteps from.
+        num_inference_steps (`int`):
+            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
+            must be `None`.
+        device (`str` or `torch.device`, *optional*):
+            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
+        timesteps (`List[int]`, *optional*):
+            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
+            `num_inference_steps` and `sigmas` must be `None`.
+        sigmas (`List[float]`, *optional*):
+            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
+            `num_inference_steps` and `timesteps` must be `None`.
+
+    Returns:
+        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
+        second element is the number of inference steps.
+    """
+    if timesteps is not None and sigmas is not None:
+        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
+    if timesteps is not None:
+        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accepts_timesteps:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" timestep schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    elif sigmas is not None:
+        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accept_sigmas:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" sigmas schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    else:
+        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+    return timesteps, num_inference_steps
+
+
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
+def retrieve_latents(
+    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
+):
+    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
+        return encoder_output.latent_dist.sample(generator)
+    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
+        return encoder_output.latent_dist.mode()
+    elif hasattr(encoder_output, "latents"):
+        return encoder_output.latents
+    else:
+        raise AttributeError("Could not access latents of provided encoder_output")
+
+
+def calculate_dimensions(target_area, ratio):
+    width = math.sqrt(target_area * ratio)
+    height = width / ratio
+
+    width = round(width / 32) * 32
+    height = round(height / 32) * 32
+
+    return width, height, None
+
+
+class QwenImageMixin:
+    @property
+    def guidance_scale(self):
+        return self._guidance_scale
+
+    @property
+    def attention_kwargs(self):
+        return self._attention_kwargs
+
+    @property
+    def num_timesteps(self):
+        return self._num_timesteps
+
+    @property
+    def current_timestep(self):
+        return self._current_timestep
+
+    @property
+    def interrupt(self):
+        return self._interrupt
+
+    def enable_vae_slicing(self):
+        r"""
+        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
+        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
+        """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
+        self.vae.enable_slicing()
+
+    def disable_vae_slicing(self):
+        r"""
+        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
+        computing decoding in one step.
+        """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
+        self.vae.disable_slicing()
+
+    def enable_vae_tiling(self):
+        r"""
+        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
+        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
+        processing larger images.
+        """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
+        self.vae.enable_tiling()
+
+    def disable_vae_tiling(self):
+        r"""
+        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
+        computing decoding in one step.
+        """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
+        self.vae.disable_tiling()
+
+    def _extract_masked_hidden(self, hidden_states: torch.Tensor, mask: torch.Tensor):
+        bool_mask = mask.bool()
+        valid_lengths = bool_mask.sum(dim=1)
+        selected = hidden_states[bool_mask]
+        split_result = torch.split(selected, valid_lengths.tolist(), dim=0)
+
+        return split_result
+
+    def _get_qwen_prompt_embeds(
+        self,
+        prompt: Union[str, List[str]],
+        image: Optional[torch.Tensor] = None,
+        device: Optional[torch.device] = None,
+        dtype: Optional[torch.dtype] = None,
+    ):
+        device = device or self._execution_device
+        dtype = dtype or self.text_encoder.dtype
+        prompt = [prompt] if isinstance(prompt, str) else prompt
+
+        template = self.prompt_template_encode
+        drop_idx = self.prompt_template_encode_start_idx
+        txt = [template.format(e) for e in prompt]
+        use_multimodal = image is not None and hasattr(self, "processor")
+
+        if use_multimodal:
+            # --- Multimodal (text+image) ---
+            model_inputs = self.processor(
+                text=txt,
+                images=image,
+                padding=True,
+                return_tensors="pt",
+            ).to(device)
+
+            outputs = self.text_encoder(
+                input_ids=model_inputs.input_ids,
+                attention_mask=model_inputs.attention_mask,
+                pixel_values=model_inputs.pixel_values,
+                image_grid_thw=model_inputs.image_grid_thw,
+                output_hidden_states=True,
+            )
+            hidden_states = outputs.hidden_states[-1]
+            attn_mask = model_inputs.attention_mask
+        else:
+            # --- Text-only ---
+            txt_tokens = self.tokenizer(
+                txt,
+                max_length=self.tokenizer_max_length + drop_idx,
+                padding=True,
+                truncation=True,
+                return_tensors="pt",
+            ).to(device)
+
+            outputs = self.text_encoder(
+                input_ids=txt_tokens.input_ids,
+                attention_mask=txt_tokens.attention_mask,
+                output_hidden_states=True,
+            )
+            hidden_states = outputs.hidden_states[-1]
+            attn_mask = txt_tokens.attention_mask
+
+        split_hidden_states = self._extract_masked_hidden(hidden_states, attn_mask)
+        split_hidden_states = [e[drop_idx:] for e in split_hidden_states]
+
+        attn_mask_list = [torch.ones(e.size(0), dtype=torch.long, device=e.device) for e in split_hidden_states]
+        max_seq_len = max(e.size(0) for e in split_hidden_states)
+
+        prompt_embeds = torch.stack(
+            [torch.cat([u, u.new_zeros(max_seq_len - u.size(0), u.size(1))]) for u in split_hidden_states]
+        )
+        encoder_attention_mask = torch.stack(
+            [torch.cat([u, u.new_zeros(max_seq_len - u.size(0))]) for u in attn_mask_list]
+        )
+
+        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
+        return prompt_embeds, encoder_attention_mask
+
+    @staticmethod
+    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
+        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
+        latents = latents.permute(0, 2, 4, 1, 3, 5)
+        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
+
+        return latents
+
+    @staticmethod
+    def _unpack_latents(latents, height, width, vae_scale_factor):
+        batch_size, num_patches, channels = latents.shape
+
+        # VAE applies 8x compression on images but we must also account for packing which requires
+        # latent height and width to be divisible by 2.
+        height = 2 * (int(height) // (vae_scale_factor * 2))
+        width = 2 * (int(width) // (vae_scale_factor * 2))
+
+        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
+        latents = latents.permute(0, 3, 1, 4, 2, 5)
+
+        latents = latents.reshape(batch_size, channels // (2 * 2), 1, height, width)
+
+        return latents
+
+
+class QwenImagePipelineMixin(QwenImageMixin):
+    def encode_prompt(
+        self,
+        prompt: Union[str, List[str]],
+        device: Optional[torch.device] = None,
+        num_images_per_prompt: int = 1,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        prompt_embeds_mask: Optional[torch.Tensor] = None,
+        max_sequence_length: int = 1024,
+    ):
+        r"""
+
+        Args:
+            prompt (`str` or `List[str]`, *optional*):
+                prompt to be encoded
+            device: (`torch.device`):
+                torch device
+            num_images_per_prompt (`int`):
+                number of images that should be generated per prompt
+            prompt_embeds (`torch.Tensor`, *optional*):
+                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
+                provided, text embeddings will be generated from `prompt` input argument.
+            prompt_embeds_mask (`torch.Tensor`, *optional*): Pre-generation masks.
+            max_sequence_length (`int`): Maximum sequence length to use.
+        """
+        device = device or self._execution_device
+
+        prompt = [prompt] if isinstance(prompt, str) else prompt
+        batch_size = len(prompt) if prompt_embeds is None else prompt_embeds.shape[0]
+
+        if prompt_embeds is None:
+            prompt_embeds, prompt_embeds_mask = self._get_qwen_prompt_embeds(prompt, device)
+
+        prompt_embeds = prompt_embeds[:, :max_sequence_length]
+        prompt_embeds_mask = prompt_embeds_mask[:, :max_sequence_length]
+
+        _, seq_len, _ = prompt_embeds.shape
+        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
+        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
+        prompt_embeds_mask = prompt_embeds_mask.repeat(1, num_images_per_prompt, 1)
+        prompt_embeds_mask = prompt_embeds_mask.view(batch_size * num_images_per_prompt, seq_len)
+
+        return prompt_embeds, prompt_embeds_mask
+
+
+class QwenImageEditPipelineMixin(QwenImageMixin):
+    def encode_prompt(
+        self,
+        prompt: Union[str, List[str]],
+        image: Optional[torch.Tensor] = None,
+        device: Optional[torch.device] = None,
+        num_images_per_prompt: int = 1,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        prompt_embeds_mask: Optional[torch.Tensor] = None,
+        max_sequence_length: int = 1024,
+    ):
+        r"""
+
+        Args:
+            prompt (`str` or `List[str]`, *optional*):
+                prompt to be encoded
+            image (`torch.Tensor`, *optional*):
+                image to be encoded
+            device: (`torch.device`):
+                torch device
+            num_images_per_prompt (`int`):
+                number of images that should be generated per prompt
+            prompt_embeds (`torch.Tensor`, *optional*):
+                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
+                provided, text embeddings will be generated from `prompt` input argument.
+        """
+        device = device or self._execution_device
+
+        prompt = [prompt] if isinstance(prompt, str) else prompt
+        batch_size = len(prompt) if prompt_embeds is None else prompt_embeds.shape[0]
+
+        if prompt_embeds is None:
+            prompt_embeds, prompt_embeds_mask = self._get_qwen_prompt_embeds(prompt, image, device)
+
+        _, seq_len, _ = prompt_embeds.shape
+        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
+        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
+        prompt_embeds_mask = prompt_embeds_mask.repeat(1, num_images_per_prompt, 1)
+        prompt_embeds_mask = prompt_embeds_mask.view(batch_size * num_images_per_prompt, seq_len)
+
+        return prompt_embeds, prompt_embeds_mask
+
+
+class QwenImageEditPlusPipelineMixin(QwenImageEditPipelineMixin):
+    def _get_qwen_prompt_embeds(
+        self,
+        prompt: Union[str, List[str]] = None,
+        image: Optional[torch.Tensor] = None,
+        device: Optional[torch.device] = None,
+        dtype: Optional[torch.dtype] = None,
+    ):
+        device = device or self._execution_device
+        dtype = dtype or self.text_encoder.dtype
+
+        prompt = [prompt] if isinstance(prompt, str) else prompt
+        img_prompt_template = "Picture {}: <|vision_start|><|image_pad|><|vision_end|>"
+        if isinstance(image, list):
+            base_img_prompt = ""
+            for i, img in enumerate(image):
+                base_img_prompt += img_prompt_template.format(i + 1)
+        elif image is not None:
+            base_img_prompt = img_prompt_template.format(1)
+        else:
+            base_img_prompt = ""
+
+        template = self.prompt_template_encode
+
+        drop_idx = self.prompt_template_encode_start_idx
+        txt = [template.format(base_img_prompt + e) for e in prompt]
+
+        model_inputs = self.processor(
+            text=txt,
+            images=image,
+            padding=True,
+            return_tensors="pt",
+        ).to(device)
+
+        outputs = self.text_encoder(
+            input_ids=model_inputs.input_ids,
+            attention_mask=model_inputs.attention_mask,
+            pixel_values=model_inputs.pixel_values,
+            image_grid_thw=model_inputs.image_grid_thw,
+            output_hidden_states=True,
+        )
+
+        hidden_states = outputs.hidden_states[-1]
+        split_hidden_states = self._extract_masked_hidden(hidden_states, model_inputs.attention_mask)
+        split_hidden_states = [e[drop_idx:] for e in split_hidden_states]
+        attn_mask_list = [torch.ones(e.size(0), dtype=torch.long, device=e.device) for e in split_hidden_states]
+        max_seq_len = max([e.size(0) for e in split_hidden_states])
+        prompt_embeds = torch.stack(
+            [torch.cat([u, u.new_zeros(max_seq_len - u.size(0), u.size(1))]) for u in split_hidden_states]
+        )
+        encoder_attention_mask = torch.stack(
+            [torch.cat([u, u.new_zeros(max_seq_len - u.size(0))]) for u in attn_mask_list]
+        )
+
+        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
+
+        return prompt_embeds, encoder_attention_mask
diff --git a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage.py b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage.py
index 33dc2039b986..822f1af69ffe 100644
--- a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage.py
+++ b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -23,10 +22,11 @@
 from ...loaders import QwenImageLoraLoaderMixin
 from ...models import AutoencoderKLQwenImage, QwenImageTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import deprecate, is_torch_xla_available, logging, replace_example_docstring
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
 from .pipeline_output import QwenImagePipelineOutput
+from .pipeline_qwen_utils import QwenImagePipelineMixin, calculate_shift, retrieve_timesteps
 
 
 if is_torch_xla_available():
@@ -56,80 +56,7 @@
 """
 
 
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
-class QwenImagePipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
+class QwenImagePipeline(DiffusionPipeline, QwenImagePipelineMixin, QwenImageLoraLoaderMixin):
     r"""
     The QwenImage pipeline for text-to-image generation.
 
@@ -343,59 +270,6 @@ def _unpack_latents(latents, height, width, vae_scale_factor):
 
         return latents
 
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
-        deprecate(
-            "enable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_slicing()
-
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
-        deprecate(
-            "disable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_slicing()
-
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
-        deprecate(
-            "enable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_tiling()
-
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
-        deprecate(
-            "disable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_tiling()
-
     def prepare_latents(
         self,
         batch_size,
@@ -428,26 +302,6 @@ def prepare_latents(
 
         return latents
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def attention_kwargs(self):
-        return self._attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def current_timestep(self):
-        return self._current_timestep
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet.py b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet.py
index 5111096d93c1..cec81f72dacb 100644
--- a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet.py
+++ b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -24,10 +23,11 @@
 from ...models import AutoencoderKLQwenImage, QwenImageTransformer2DModel
 from ...models.controlnets.controlnet_qwenimage import QwenImageControlNetModel, QwenImageMultiControlNetModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import deprecate, is_torch_xla_available, logging, replace_example_docstring
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
 from .pipeline_output import QwenImagePipelineOutput
+from .pipeline_qwen_utils import QwenImagePipelineMixin, calculate_shift, retrieve_latents, retrieve_timesteps
 
 
 if is_torch_xla_available():
@@ -101,95 +101,7 @@
 """
 
 
-# Coped from diffusers.pipelines.qwenimage.pipeline_qwenimage.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
-class QwenImageControlNetPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
+class QwenImageControlNetPipeline(DiffusionPipeline, QwenImagePipelineMixin, QwenImageLoraLoaderMixin):
     r"""
     The QwenImage pipeline for text-to-image generation.
 
@@ -239,93 +151,6 @@ def __init__(
         self.prompt_template_encode_start_idx = 34
         self.default_sample_size = 128
 
-    # Coped from diffusers.pipelines.qwenimage.pipeline_qwenimage.extract_masked_hidden
-    def _extract_masked_hidden(self, hidden_states: torch.Tensor, mask: torch.Tensor):
-        bool_mask = mask.bool()
-        valid_lengths = bool_mask.sum(dim=1)
-        selected = hidden_states[bool_mask]
-        split_result = torch.split(selected, valid_lengths.tolist(), dim=0)
-
-        return split_result
-
-    # Coped from diffusers.pipelines.qwenimage.pipeline_qwenimage.get_qwen_prompt_embeds
-    def _get_qwen_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        template = self.prompt_template_encode
-        drop_idx = self.prompt_template_encode_start_idx
-        txt = [template.format(e) for e in prompt]
-        txt_tokens = self.tokenizer(
-            txt, max_length=self.tokenizer_max_length + drop_idx, padding=True, truncation=True, return_tensors="pt"
-        ).to(device)
-        encoder_hidden_states = self.text_encoder(
-            input_ids=txt_tokens.input_ids,
-            attention_mask=txt_tokens.attention_mask,
-            output_hidden_states=True,
-        )
-        hidden_states = encoder_hidden_states.hidden_states[-1]
-        split_hidden_states = self._extract_masked_hidden(hidden_states, txt_tokens.attention_mask)
-        split_hidden_states = [e[drop_idx:] for e in split_hidden_states]
-        attn_mask_list = [torch.ones(e.size(0), dtype=torch.long, device=e.device) for e in split_hidden_states]
-        max_seq_len = max([e.size(0) for e in split_hidden_states])
-        prompt_embeds = torch.stack(
-            [torch.cat([u, u.new_zeros(max_seq_len - u.size(0), u.size(1))]) for u in split_hidden_states]
-        )
-        encoder_attention_mask = torch.stack(
-            [torch.cat([u, u.new_zeros(max_seq_len - u.size(0))]) for u in attn_mask_list]
-        )
-
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        return prompt_embeds, encoder_attention_mask
-
-    # Coped from diffusers.pipelines.qwenimage.pipeline_qwenimage.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        prompt_embeds_mask: Optional[torch.Tensor] = None,
-        max_sequence_length: int = 1024,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-        """
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt) if prompt_embeds is None else prompt_embeds.shape[0]
-
-        if prompt_embeds is None:
-            prompt_embeds, prompt_embeds_mask = self._get_qwen_prompt_embeds(prompt, device)
-
-        _, seq_len, _ = prompt_embeds.shape
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-        prompt_embeds_mask = prompt_embeds_mask.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds_mask = prompt_embeds_mask.view(batch_size * num_images_per_prompt, seq_len)
-
-        return prompt_embeds, prompt_embeds_mask
-
     def check_inputs(
         self,
         prompt,
@@ -381,85 +206,6 @@ def check_inputs(
         if max_sequence_length is not None and max_sequence_length > 1024:
             raise ValueError(f"`max_sequence_length` cannot be greater than 1024 but is {max_sequence_length}")
 
-    @staticmethod
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._pack_latents
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._unpack_latents
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), 1, height, width)
-
-        return latents
-
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
-        deprecate(
-            "enable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_slicing()
-
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
-        deprecate(
-            "disable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_slicing()
-
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
-        deprecate(
-            "enable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_tiling()
-
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
-        deprecate(
-            "disable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_tiling()
-
     # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline.prepare_latents
     def prepare_latents(
         self,
@@ -528,26 +274,6 @@ def prepare_image(
 
         return image
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def attention_kwargs(self):
-        return self._attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def current_timestep(self):
-        return self._current_timestep
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet_inpaint.py b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet_inpaint.py
index 102a813ab582..6b06392fd66c 100644
--- a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet_inpaint.py
+++ b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet_inpaint.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -28,6 +27,7 @@
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
 from .pipeline_output import QwenImagePipelineOutput
+from .pipeline_qwen_utils import QwenImagePipelineMixin, calculate_shift, retrieve_timesteps
 
 
 if is_torch_xla_available():
@@ -74,95 +74,7 @@
 """
 
 
-# Coped from diffusers.pipelines.qwenimage.pipeline_qwenimage.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
-class QwenImageControlNetInpaintPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
+class QwenImageControlNetInpaintPipeline(DiffusionPipeline, QwenImagePipelineMixin, QwenImageLoraLoaderMixin):
     r"""
     The QwenImage pipeline for text-to-image generation.
 
@@ -221,92 +133,6 @@ def __init__(
         self.prompt_template_encode_start_idx = 34
         self.default_sample_size = 128
 
-    # Coped from diffusers.pipelines.qwenimage.pipeline_qwenimage.extract_masked_hidden
-    def _extract_masked_hidden(self, hidden_states: torch.Tensor, mask: torch.Tensor):
-        bool_mask = mask.bool()
-        valid_lengths = bool_mask.sum(dim=1)
-        selected = hidden_states[bool_mask]
-        split_result = torch.split(selected, valid_lengths.tolist(), dim=0)
-
-        return split_result
-
-    # Coped from diffusers.pipelines.qwenimage.pipeline_qwenimage.get_qwen_prompt_embeds
-    def _get_qwen_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        template = self.prompt_template_encode
-        drop_idx = self.prompt_template_encode_start_idx
-        txt = [template.format(e) for e in prompt]
-        txt_tokens = self.tokenizer(
-            txt, max_length=self.tokenizer_max_length + drop_idx, padding=True, truncation=True, return_tensors="pt"
-        ).to(self.device)
-        encoder_hidden_states = self.text_encoder(
-            input_ids=txt_tokens.input_ids,
-            attention_mask=txt_tokens.attention_mask,
-            output_hidden_states=True,
-        )
-        hidden_states = encoder_hidden_states.hidden_states[-1]
-        split_hidden_states = self._extract_masked_hidden(hidden_states, txt_tokens.attention_mask)
-        split_hidden_states = [e[drop_idx:] for e in split_hidden_states]
-        attn_mask_list = [torch.ones(e.size(0), dtype=torch.long, device=e.device) for e in split_hidden_states]
-        max_seq_len = max([e.size(0) for e in split_hidden_states])
-        prompt_embeds = torch.stack(
-            [torch.cat([u, u.new_zeros(max_seq_len - u.size(0), u.size(1))]) for u in split_hidden_states]
-        )
-        encoder_attention_mask = torch.stack(
-            [torch.cat([u, u.new_zeros(max_seq_len - u.size(0))]) for u in attn_mask_list]
-        )
-
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        return prompt_embeds, encoder_attention_mask
-
-    # Coped from diffusers.pipelines.qwenimage.pipeline_qwenimage.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        prompt_embeds_mask: Optional[torch.Tensor] = None,
-        max_sequence_length: int = 1024,
-    ):
-        r"""
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-        """
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt) if prompt_embeds is None else prompt_embeds.shape[0]
-
-        if prompt_embeds is None:
-            prompt_embeds, prompt_embeds_mask = self._get_qwen_prompt_embeds(prompt, device)
-
-        _, seq_len, _ = prompt_embeds.shape
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-        prompt_embeds_mask = prompt_embeds_mask.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds_mask = prompt_embeds_mask.view(batch_size * num_images_per_prompt, seq_len)
-
-        return prompt_embeds, prompt_embeds_mask
-
     def check_inputs(
         self,
         prompt,
@@ -362,61 +188,6 @@ def check_inputs(
         if max_sequence_length is not None and max_sequence_length > 1024:
             raise ValueError(f"`max_sequence_length` cannot be greater than 1024 but is {max_sequence_length}")
 
-    @staticmethod
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._pack_latents
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._unpack_latents
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), 1, height, width)
-
-        return latents
-
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        self.vae.enable_tiling()
-
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_tiling()
-
     # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline.prepare_latents
     def prepare_latents(
         self,
@@ -568,26 +339,6 @@ def prepare_image_with_mask(
 
         return control_image
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def attention_kwargs(self):
-        return self._attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def current_timestep(self):
-        return self._current_timestep
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit.py b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit.py
index ed37b238c8c9..f246ed277316 100644
--- a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit.py
+++ b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit.py
@@ -12,8 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
-import math
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -24,10 +22,17 @@
 from ...loaders import QwenImageLoraLoaderMixin
 from ...models import AutoencoderKLQwenImage, QwenImageTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import deprecate, is_torch_xla_available, logging, replace_example_docstring
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
 from .pipeline_output import QwenImagePipelineOutput
+from .pipeline_qwen_utils import (
+    QwenImageEditPipelineMixin,
+    calculate_dimensions,
+    calculate_shift,
+    retrieve_latents,
+    retrieve_timesteps,
+)
 
 
 if is_torch_xla_available():
@@ -64,105 +69,7 @@
 """
 
 
-# Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
-def calculate_dimensions(target_area, ratio):
-    width = math.sqrt(target_area * ratio)
-    height = width / ratio
-
-    width = round(width / 32) * 32
-    height = round(height / 32) * 32
-
-    return width, height, None
-
-
-class QwenImageEditPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
+class QwenImageEditPipeline(DiffusionPipeline, QwenImageEditPipelineMixin, QwenImageLoraLoaderMixin):
     r"""
     The Qwen-Image-Edit pipeline for image editing.
 
@@ -214,103 +121,6 @@ def __init__(
         self.prompt_template_encode_start_idx = 64
         self.default_sample_size = 128
 
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._extract_masked_hidden
-    def _extract_masked_hidden(self, hidden_states: torch.Tensor, mask: torch.Tensor):
-        bool_mask = mask.bool()
-        valid_lengths = bool_mask.sum(dim=1)
-        selected = hidden_states[bool_mask]
-        split_result = torch.split(selected, valid_lengths.tolist(), dim=0)
-
-        return split_result
-
-    def _get_qwen_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        image: Optional[torch.Tensor] = None,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        template = self.prompt_template_encode
-        drop_idx = self.prompt_template_encode_start_idx
-        txt = [template.format(e) for e in prompt]
-
-        model_inputs = self.processor(
-            text=txt,
-            images=image,
-            padding=True,
-            return_tensors="pt",
-        ).to(device)
-
-        outputs = self.text_encoder(
-            input_ids=model_inputs.input_ids,
-            attention_mask=model_inputs.attention_mask,
-            pixel_values=model_inputs.pixel_values,
-            image_grid_thw=model_inputs.image_grid_thw,
-            output_hidden_states=True,
-        )
-
-        hidden_states = outputs.hidden_states[-1]
-        split_hidden_states = self._extract_masked_hidden(hidden_states, model_inputs.attention_mask)
-        split_hidden_states = [e[drop_idx:] for e in split_hidden_states]
-        attn_mask_list = [torch.ones(e.size(0), dtype=torch.long, device=e.device) for e in split_hidden_states]
-        max_seq_len = max([e.size(0) for e in split_hidden_states])
-        prompt_embeds = torch.stack(
-            [torch.cat([u, u.new_zeros(max_seq_len - u.size(0), u.size(1))]) for u in split_hidden_states]
-        )
-        encoder_attention_mask = torch.stack(
-            [torch.cat([u, u.new_zeros(max_seq_len - u.size(0))]) for u in attn_mask_list]
-        )
-
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        return prompt_embeds, encoder_attention_mask
-
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        image: Optional[torch.Tensor] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        prompt_embeds_mask: Optional[torch.Tensor] = None,
-        max_sequence_length: int = 1024,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            image (`torch.Tensor`, *optional*):
-                image to be encoded
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-        """
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt) if prompt_embeds is None else prompt_embeds.shape[0]
-
-        if prompt_embeds is None:
-            prompt_embeds, prompt_embeds_mask = self._get_qwen_prompt_embeds(prompt, image, device)
-
-        _, seq_len, _ = prompt_embeds.shape
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-        prompt_embeds_mask = prompt_embeds_mask.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds_mask = prompt_embeds_mask.view(batch_size * num_images_per_prompt, seq_len)
-
-        return prompt_embeds, prompt_embeds_mask
-
     def check_inputs(
         self,
         prompt,
@@ -366,32 +176,6 @@ def check_inputs(
         if max_sequence_length is not None and max_sequence_length > 1024:
             raise ValueError(f"`max_sequence_length` cannot be greater than 1024 but is {max_sequence_length}")
 
-    @staticmethod
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._pack_latents
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._unpack_latents
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), 1, height, width)
-
-        return latents
-
     def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
         if isinstance(generator, list):
             image_latents = [
@@ -415,59 +199,6 @@ def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
 
         return image_latents
 
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
-        deprecate(
-            "enable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_slicing()
-
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
-        deprecate(
-            "disable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_slicing()
-
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
-        deprecate(
-            "enable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_tiling()
-
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
-        deprecate(
-            "disable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_tiling()
-
     def prepare_latents(
         self,
         image,
@@ -523,26 +254,6 @@ def prepare_latents(
 
         return latents, image_latents
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def attention_kwargs(self):
-        return self._attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def current_timestep(self):
-        return self._current_timestep
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit_inpaint.py b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit_inpaint.py
index d54d1881fa4e..9e721d030761 100644
--- a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit_inpaint.py
+++ b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit_inpaint.py
@@ -12,8 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
-import math
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -25,10 +23,17 @@
 from ...loaders import QwenImageLoraLoaderMixin
 from ...models import AutoencoderKLQwenImage, QwenImageTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import deprecate, is_torch_xla_available, logging, replace_example_docstring
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
 from .pipeline_output import QwenImagePipelineOutput
+from .pipeline_qwen_utils import (
+    QwenImageEditPipelineMixin,
+    calculate_dimensions,
+    calculate_shift,
+    retrieve_latents,
+    retrieve_timesteps,
+)
 
 
 if is_torch_xla_available():
@@ -65,106 +70,7 @@
 """
 
 
-# Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
-# Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage_edit.calculate_dimensions
-def calculate_dimensions(target_area, ratio):
-    width = math.sqrt(target_area * ratio)
-    height = width / ratio
-
-    width = round(width / 32) * 32
-    height = round(height / 32) * 32
-
-    return width, height, None
-
-
-class QwenImageEditInpaintPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
+class QwenImageEditInpaintPipeline(DiffusionPipeline, QwenImageEditPipelineMixin, QwenImageLoraLoaderMixin):
     r"""
     The Qwen-Image-Edit pipeline for image editing.
 
@@ -224,105 +130,6 @@ def __init__(
         self.prompt_template_encode_start_idx = 64
         self.default_sample_size = 128
 
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._extract_masked_hidden
-    def _extract_masked_hidden(self, hidden_states: torch.Tensor, mask: torch.Tensor):
-        bool_mask = mask.bool()
-        valid_lengths = bool_mask.sum(dim=1)
-        selected = hidden_states[bool_mask]
-        split_result = torch.split(selected, valid_lengths.tolist(), dim=0)
-
-        return split_result
-
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage_edit.QwenImageEditPipeline._get_qwen_prompt_embeds
-    def _get_qwen_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        image: Optional[torch.Tensor] = None,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        template = self.prompt_template_encode
-        drop_idx = self.prompt_template_encode_start_idx
-        txt = [template.format(e) for e in prompt]
-
-        model_inputs = self.processor(
-            text=txt,
-            images=image,
-            padding=True,
-            return_tensors="pt",
-        ).to(device)
-
-        outputs = self.text_encoder(
-            input_ids=model_inputs.input_ids,
-            attention_mask=model_inputs.attention_mask,
-            pixel_values=model_inputs.pixel_values,
-            image_grid_thw=model_inputs.image_grid_thw,
-            output_hidden_states=True,
-        )
-
-        hidden_states = outputs.hidden_states[-1]
-        split_hidden_states = self._extract_masked_hidden(hidden_states, model_inputs.attention_mask)
-        split_hidden_states = [e[drop_idx:] for e in split_hidden_states]
-        attn_mask_list = [torch.ones(e.size(0), dtype=torch.long, device=e.device) for e in split_hidden_states]
-        max_seq_len = max([e.size(0) for e in split_hidden_states])
-        prompt_embeds = torch.stack(
-            [torch.cat([u, u.new_zeros(max_seq_len - u.size(0), u.size(1))]) for u in split_hidden_states]
-        )
-        encoder_attention_mask = torch.stack(
-            [torch.cat([u, u.new_zeros(max_seq_len - u.size(0))]) for u in attn_mask_list]
-        )
-
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        return prompt_embeds, encoder_attention_mask
-
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage_edit.QwenImageEditPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        image: Optional[torch.Tensor] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        prompt_embeds_mask: Optional[torch.Tensor] = None,
-        max_sequence_length: int = 1024,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            image (`torch.Tensor`, *optional*):
-                image to be encoded
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-        """
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt) if prompt_embeds is None else prompt_embeds.shape[0]
-
-        if prompt_embeds is None:
-            prompt_embeds, prompt_embeds_mask = self._get_qwen_prompt_embeds(prompt, image, device)
-
-        _, seq_len, _ = prompt_embeds.shape
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-        prompt_embeds_mask = prompt_embeds_mask.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds_mask = prompt_embeds_mask.view(batch_size * num_images_per_prompt, seq_len)
-
-        return prompt_embeds, prompt_embeds_mask
-
     # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage_inpaint.QwenImageInpaintPipeline.check_inputs
     def check_inputs(
         self,
@@ -399,32 +206,6 @@ def check_inputs(
         if max_sequence_length is not None and max_sequence_length > 1024:
             raise ValueError(f"`max_sequence_length` cannot be greater than 1024 but is {max_sequence_length}")
 
-    @staticmethod
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._pack_latents
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._unpack_latents
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), 1, height, width)
-
-        return latents
-
     # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage_img2img.QwenImageImg2ImgPipeline._encode_vae_image
     def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
         if isinstance(generator, list):
@@ -461,59 +242,6 @@ def get_timesteps(self, num_inference_steps, strength, device):
 
         return timesteps, num_inference_steps - t_start
 
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
-        deprecate(
-            "enable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_slicing()
-
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
-        deprecate(
-            "disable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_slicing()
-
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
-        deprecate(
-            "enable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_tiling()
-
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
-        deprecate(
-            "disable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_tiling()
-
     # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage_inpaint.QwenImageInpaintPipeline.prepare_latents
     def prepare_latents(
         self,
@@ -656,26 +384,6 @@ def prepare_mask_latents(
 
         return mask, masked_image_latents
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def attention_kwargs(self):
-        return self._attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def current_timestep(self):
-        return self._current_timestep
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit_plus.py b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit_plus.py
index ec203edf166c..07963c09a32c 100644
--- a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit_plus.py
+++ b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit_plus.py
@@ -12,8 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
-import math
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -28,6 +26,13 @@
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
 from .pipeline_output import QwenImagePipelineOutput
+from .pipeline_qwen_utils import (
+    QwenImageEditPlusPipelineMixin,
+    calculate_dimensions,
+    calculate_shift,
+    retrieve_latents,
+    retrieve_timesteps,
+)
 
 
 if is_torch_xla_available():
@@ -67,105 +72,7 @@
 VAE_IMAGE_SIZE = 1024 * 1024
 
 
-# Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
-def calculate_dimensions(target_area, ratio):
-    width = math.sqrt(target_area * ratio)
-    height = width / ratio
-
-    width = round(width / 32) * 32
-    height = round(height / 32) * 32
-
-    return width, height
-
-
-class QwenImageEditPlusPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
+class QwenImageEditPlusPipeline(DiffusionPipeline, QwenImageEditPlusPipelineMixin, QwenImageLoraLoaderMixin):
     r"""
     The Qwen-Image-Edit pipeline for image editing.
 
@@ -217,114 +124,6 @@ def __init__(
         self.prompt_template_encode_start_idx = 64
         self.default_sample_size = 128
 
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._extract_masked_hidden
-    def _extract_masked_hidden(self, hidden_states: torch.Tensor, mask: torch.Tensor):
-        bool_mask = mask.bool()
-        valid_lengths = bool_mask.sum(dim=1)
-        selected = hidden_states[bool_mask]
-        split_result = torch.split(selected, valid_lengths.tolist(), dim=0)
-
-        return split_result
-
-    def _get_qwen_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        image: Optional[torch.Tensor] = None,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        img_prompt_template = "Picture {}: <|vision_start|><|image_pad|><|vision_end|>"
-        if isinstance(image, list):
-            base_img_prompt = ""
-            for i, img in enumerate(image):
-                base_img_prompt += img_prompt_template.format(i + 1)
-        elif image is not None:
-            base_img_prompt = img_prompt_template.format(1)
-        else:
-            base_img_prompt = ""
-
-        template = self.prompt_template_encode
-
-        drop_idx = self.prompt_template_encode_start_idx
-        txt = [template.format(base_img_prompt + e) for e in prompt]
-
-        model_inputs = self.processor(
-            text=txt,
-            images=image,
-            padding=True,
-            return_tensors="pt",
-        ).to(device)
-
-        outputs = self.text_encoder(
-            input_ids=model_inputs.input_ids,
-            attention_mask=model_inputs.attention_mask,
-            pixel_values=model_inputs.pixel_values,
-            image_grid_thw=model_inputs.image_grid_thw,
-            output_hidden_states=True,
-        )
-
-        hidden_states = outputs.hidden_states[-1]
-        split_hidden_states = self._extract_masked_hidden(hidden_states, model_inputs.attention_mask)
-        split_hidden_states = [e[drop_idx:] for e in split_hidden_states]
-        attn_mask_list = [torch.ones(e.size(0), dtype=torch.long, device=e.device) for e in split_hidden_states]
-        max_seq_len = max([e.size(0) for e in split_hidden_states])
-        prompt_embeds = torch.stack(
-            [torch.cat([u, u.new_zeros(max_seq_len - u.size(0), u.size(1))]) for u in split_hidden_states]
-        )
-        encoder_attention_mask = torch.stack(
-            [torch.cat([u, u.new_zeros(max_seq_len - u.size(0))]) for u in attn_mask_list]
-        )
-
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        return prompt_embeds, encoder_attention_mask
-
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage_edit.QwenImageEditPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        image: Optional[torch.Tensor] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        prompt_embeds_mask: Optional[torch.Tensor] = None,
-        max_sequence_length: int = 1024,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            image (`torch.Tensor`, *optional*):
-                image to be encoded
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-        """
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt) if prompt_embeds is None else prompt_embeds.shape[0]
-
-        if prompt_embeds is None:
-            prompt_embeds, prompt_embeds_mask = self._get_qwen_prompt_embeds(prompt, image, device)
-
-        _, seq_len, _ = prompt_embeds.shape
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-        prompt_embeds_mask = prompt_embeds_mask.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds_mask = prompt_embeds_mask.view(batch_size * num_images_per_prompt, seq_len)
-
-        return prompt_embeds, prompt_embeds_mask
-
     # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage_edit.QwenImageEditPipeline.check_inputs
     def check_inputs(
         self,
@@ -381,32 +180,6 @@ def check_inputs(
         if max_sequence_length is not None and max_sequence_length > 1024:
             raise ValueError(f"`max_sequence_length` cannot be greater than 1024 but is {max_sequence_length}")
 
-    @staticmethod
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._pack_latents
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._unpack_latents
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), 1, height, width)
-
-        return latents
-
     # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage_edit.QwenImageEditPipeline._encode_vae_image
     def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
         if isinstance(generator, list):
@@ -492,26 +265,6 @@ def prepare_latents(
 
         return latents, image_latents
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def attention_kwargs(self):
-        return self._attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def current_timestep(self):
-        return self._current_timestep
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
@@ -628,7 +381,7 @@ def __call__(
             returning a tuple, the first element is a list with the generated images.
         """
         image_size = image[-1].size if isinstance(image, list) else image.size
-        calculated_width, calculated_height = calculate_dimensions(1024 * 1024, image_size[0] / image_size[1])
+        calculated_width, calculated_height, _ = calculate_dimensions(1024 * 1024, image_size[0] / image_size[1])
         height = height or calculated_height
         width = width or calculated_width
 
@@ -674,10 +427,10 @@ def __call__(
             vae_images = []
             for img in image:
                 image_width, image_height = img.size
-                condition_width, condition_height = calculate_dimensions(
+                condition_width, condition_height, _ = calculate_dimensions(
                     CONDITION_IMAGE_SIZE, image_width / image_height
                 )
-                vae_width, vae_height = calculate_dimensions(VAE_IMAGE_SIZE, image_width / image_height)
+                vae_width, vae_height, _ = calculate_dimensions(VAE_IMAGE_SIZE, image_width / image_height)
                 condition_image_sizes.append((condition_width, condition_height))
                 vae_image_sizes.append((vae_width, vae_height))
                 condition_images.append(self.image_processor.resize(img, condition_height, condition_width))
diff --git a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_img2img.py b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_img2img.py
index cb4c5d8016bb..fc0b852e8352 100644
--- a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_img2img.py
+++ b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_img2img.py
@@ -1,4 +1,3 @@
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -9,10 +8,11 @@
 from ...loaders import QwenImageLoraLoaderMixin
 from ...models import AutoencoderKLQwenImage, QwenImageTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import deprecate, is_torch_xla_available, logging, replace_example_docstring
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
 from .pipeline_output import QwenImagePipelineOutput
+from .pipeline_qwen_utils import QwenImagePipelineMixin, calculate_shift, retrieve_latents, retrieve_timesteps
 
 
 if is_torch_xla_available():
@@ -43,95 +43,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
-# Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
-class QwenImageImg2ImgPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
+class QwenImageImg2ImgPipeline(DiffusionPipeline, QwenImagePipelineMixin, QwenImageLoraLoaderMixin):
     r"""
     The QwenImage pipeline for text-to-image generation.
 
@@ -182,54 +94,6 @@ def __init__(
         self.prompt_template_encode_start_idx = 34
         self.default_sample_size = 128
 
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._extract_masked_hidden
-    def _extract_masked_hidden(self, hidden_states: torch.Tensor, mask: torch.Tensor):
-        bool_mask = mask.bool()
-        valid_lengths = bool_mask.sum(dim=1)
-        selected = hidden_states[bool_mask]
-        split_result = torch.split(selected, valid_lengths.tolist(), dim=0)
-
-        return split_result
-
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._get_qwen_prompt_embeds
-    def _get_qwen_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        template = self.prompt_template_encode
-        drop_idx = self.prompt_template_encode_start_idx
-        txt = [template.format(e) for e in prompt]
-        txt_tokens = self.tokenizer(
-            txt, max_length=self.tokenizer_max_length + drop_idx, padding=True, truncation=True, return_tensors="pt"
-        ).to(device)
-        encoder_hidden_states = self.text_encoder(
-            input_ids=txt_tokens.input_ids,
-            attention_mask=txt_tokens.attention_mask,
-            output_hidden_states=True,
-        )
-        hidden_states = encoder_hidden_states.hidden_states[-1]
-        split_hidden_states = self._extract_masked_hidden(hidden_states, txt_tokens.attention_mask)
-        split_hidden_states = [e[drop_idx:] for e in split_hidden_states]
-        attn_mask_list = [torch.ones(e.size(0), dtype=torch.long, device=e.device) for e in split_hidden_states]
-        max_seq_len = max([e.size(0) for e in split_hidden_states])
-        prompt_embeds = torch.stack(
-            [torch.cat([u, u.new_zeros(max_seq_len - u.size(0), u.size(1))]) for u in split_hidden_states]
-        )
-        encoder_attention_mask = torch.stack(
-            [torch.cat([u, u.new_zeros(max_seq_len - u.size(0))]) for u in attn_mask_list]
-        )
-
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        return prompt_embeds, encoder_attention_mask
-
     def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
         if isinstance(generator, list):
             image_latents = [
@@ -265,48 +129,6 @@ def get_timesteps(self, num_inference_steps, strength, device):
 
         return timesteps, num_inference_steps - t_start
 
-    # Copied fromCopied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        prompt_embeds_mask: Optional[torch.Tensor] = None,
-        max_sequence_length: int = 1024,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-        """
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt) if prompt_embeds is None else prompt_embeds.shape[0]
-
-        if prompt_embeds is None:
-            prompt_embeds, prompt_embeds_mask = self._get_qwen_prompt_embeds(prompt, device)
-
-        prompt_embeds = prompt_embeds[:, :max_sequence_length]
-        prompt_embeds_mask = prompt_embeds_mask[:, :max_sequence_length]
-
-        _, seq_len, _ = prompt_embeds.shape
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-        prompt_embeds_mask = prompt_embeds_mask.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds_mask = prompt_embeds_mask.view(batch_size * num_images_per_prompt, seq_len)
-
-        return prompt_embeds, prompt_embeds_mask
-
     def check_inputs(
         self,
         prompt,
@@ -366,85 +188,6 @@ def check_inputs(
         if max_sequence_length is not None and max_sequence_length > 1024:
             raise ValueError(f"`max_sequence_length` cannot be greater than 1024 but is {max_sequence_length}")
 
-    @staticmethod
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._pack_latents
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._unpack_latents
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), 1, height, width)
-
-        return latents
-
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
-        deprecate(
-            "enable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_slicing()
-
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
-        deprecate(
-            "disable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_slicing()
-
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
-        deprecate(
-            "enable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_tiling()
-
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
-        deprecate(
-            "disable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_tiling()
-
     def prepare_latents(
         self,
         image,
@@ -502,26 +245,6 @@ def prepare_latents(
 
         return latents
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def attention_kwargs(self):
-        return self._attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def current_timestep(self):
-        return self._current_timestep
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_inpaint.py b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_inpaint.py
index 1915c27eb2bb..db85ae5763ac 100644
--- a/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_inpaint.py
+++ b/src/diffusers/pipelines/qwenimage/pipeline_qwenimage_inpaint.py
@@ -1,4 +1,3 @@
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -10,10 +9,11 @@
 from ...loaders import QwenImageLoraLoaderMixin
 from ...models import AutoencoderKLQwenImage, QwenImageTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import deprecate, is_torch_xla_available, logging, replace_example_docstring
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
 from .pipeline_output import QwenImagePipelineOutput
+from .pipeline_qwen_utils import QwenImagePipelineMixin, calculate_shift, retrieve_latents, retrieve_timesteps
 
 
 if is_torch_xla_available():
@@ -46,95 +46,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
-# Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
-class QwenImageInpaintPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
+class QwenImageInpaintPipeline(DiffusionPipeline, QwenImagePipelineMixin, QwenImageLoraLoaderMixin):
     r"""
     The QwenImage pipeline for text-to-image generation.
 
@@ -192,54 +104,6 @@ def __init__(
         self.prompt_template_encode_start_idx = 34
         self.default_sample_size = 128
 
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._extract_masked_hidden
-    def _extract_masked_hidden(self, hidden_states: torch.Tensor, mask: torch.Tensor):
-        bool_mask = mask.bool()
-        valid_lengths = bool_mask.sum(dim=1)
-        selected = hidden_states[bool_mask]
-        split_result = torch.split(selected, valid_lengths.tolist(), dim=0)
-
-        return split_result
-
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._get_qwen_prompt_embeds
-    def _get_qwen_prompt_embeds(
-        self,
-        prompt: Union[str, List[str]] = None,
-        device: Optional[torch.device] = None,
-        dtype: Optional[torch.dtype] = None,
-    ):
-        device = device or self._execution_device
-        dtype = dtype or self.text_encoder.dtype
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        template = self.prompt_template_encode
-        drop_idx = self.prompt_template_encode_start_idx
-        txt = [template.format(e) for e in prompt]
-        txt_tokens = self.tokenizer(
-            txt, max_length=self.tokenizer_max_length + drop_idx, padding=True, truncation=True, return_tensors="pt"
-        ).to(device)
-        encoder_hidden_states = self.text_encoder(
-            input_ids=txt_tokens.input_ids,
-            attention_mask=txt_tokens.attention_mask,
-            output_hidden_states=True,
-        )
-        hidden_states = encoder_hidden_states.hidden_states[-1]
-        split_hidden_states = self._extract_masked_hidden(hidden_states, txt_tokens.attention_mask)
-        split_hidden_states = [e[drop_idx:] for e in split_hidden_states]
-        attn_mask_list = [torch.ones(e.size(0), dtype=torch.long, device=e.device) for e in split_hidden_states]
-        max_seq_len = max([e.size(0) for e in split_hidden_states])
-        prompt_embeds = torch.stack(
-            [torch.cat([u, u.new_zeros(max_seq_len - u.size(0), u.size(1))]) for u in split_hidden_states]
-        )
-        encoder_attention_mask = torch.stack(
-            [torch.cat([u, u.new_zeros(max_seq_len - u.size(0))]) for u in attn_mask_list]
-        )
-
-        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
-
-        return prompt_embeds, encoder_attention_mask
-
     # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage_img2img.QwenImageImg2ImgPipeline._encode_vae_image
     def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
         if isinstance(generator, list):
@@ -276,48 +140,6 @@ def get_timesteps(self, num_inference_steps, strength, device):
 
         return timesteps, num_inference_steps - t_start
 
-    # Copied fromCopied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: Union[str, List[str]],
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        prompt_embeds_mask: Optional[torch.Tensor] = None,
-        max_sequence_length: int = 1024,
-    ):
-        r"""
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-        """
-        device = device or self._execution_device
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-        batch_size = len(prompt) if prompt_embeds is None else prompt_embeds.shape[0]
-
-        if prompt_embeds is None:
-            prompt_embeds, prompt_embeds_mask = self._get_qwen_prompt_embeds(prompt, device)
-
-        prompt_embeds = prompt_embeds[:, :max_sequence_length]
-        prompt_embeds_mask = prompt_embeds_mask[:, :max_sequence_length]
-
-        _, seq_len, _ = prompt_embeds.shape
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-        prompt_embeds_mask = prompt_embeds_mask.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds_mask = prompt_embeds_mask.view(batch_size * num_images_per_prompt, seq_len)
-
-        return prompt_embeds, prompt_embeds_mask
-
     def check_inputs(
         self,
         prompt,
@@ -393,85 +215,6 @@ def check_inputs(
         if max_sequence_length is not None and max_sequence_length > 1024:
             raise ValueError(f"`max_sequence_length` cannot be greater than 1024 but is {max_sequence_length}")
 
-    @staticmethod
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._pack_latents
-    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
-        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
-        latents = latents.permute(0, 2, 4, 1, 3, 5)
-        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
-
-        return latents
-
-    @staticmethod
-    # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline._unpack_latents
-    def _unpack_latents(latents, height, width, vae_scale_factor):
-        batch_size, num_patches, channels = latents.shape
-
-        # VAE applies 8x compression on images but we must also account for packing which requires
-        # latent height and width to be divisible by 2.
-        height = 2 * (int(height) // (vae_scale_factor * 2))
-        width = 2 * (int(width) // (vae_scale_factor * 2))
-
-        latents = latents.view(batch_size, height // 2, width // 2, channels // 4, 2, 2)
-        latents = latents.permute(0, 3, 1, 4, 2, 5)
-
-        latents = latents.reshape(batch_size, channels // (2 * 2), 1, height, width)
-
-        return latents
-
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
-        deprecate(
-            "enable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_slicing()
-
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
-        deprecate(
-            "disable_vae_slicing",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_slicing()
-
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
-        deprecate(
-            "enable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.enable_tiling()
-
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
-        deprecate(
-            "disable_vae_tiling",
-            "0.40.0",
-            depr_message,
-        )
-        self.vae.disable_tiling()
-
     def prepare_latents(
         self,
         image,
@@ -612,26 +355,6 @@ def prepare_mask_latents(
 
         return mask, masked_image_latents
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def attention_kwargs(self):
-        return self._attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def current_timestep(self):
-        return self._current_timestep
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/sana/pipeline_sana.py b/src/diffusers/pipelines/sana/pipeline_sana.py
index 2beff802c6e0..920479f34f10 100644
--- a/src/diffusers/pipelines/sana/pipeline_sana.py
+++ b/src/diffusers/pipelines/sana/pipeline_sana.py
@@ -127,7 +127,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -461,7 +461,7 @@ def encode_prompt(
 
         return prompt_embeds, prompt_attention_mask, negative_prompt_embeds, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/sana/pipeline_sana_controlnet.py b/src/diffusers/pipelines/sana/pipeline_sana_controlnet.py
index 55ed7b84ebdf..b3569047438a 100644
--- a/src/diffusers/pipelines/sana/pipeline_sana_controlnet.py
+++ b/src/diffusers/pipelines/sana/pipeline_sana_controlnet.py
@@ -134,7 +134,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -476,7 +476,7 @@ def encode_prompt(
 
         return prompt_embeds, prompt_attention_mask, negative_prompt_embeds, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/sana/pipeline_sana_sprint.py b/src/diffusers/pipelines/sana/pipeline_sana_sprint.py
index 04f45f817efb..9f6a80e3f8c7 100644
--- a/src/diffusers/pipelines/sana/pipeline_sana_sprint.py
+++ b/src/diffusers/pipelines/sana/pipeline_sana_sprint.py
@@ -78,7 +78,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -368,7 +368,7 @@ def encode_prompt(
 
         return prompt_embeds, prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/sana/pipeline_sana_sprint_img2img.py b/src/diffusers/pipelines/sana/pipeline_sana_sprint_img2img.py
index 8899ed84c4e5..a9cbf65660f8 100644
--- a/src/diffusers/pipelines/sana/pipeline_sana_sprint_img2img.py
+++ b/src/diffusers/pipelines/sana/pipeline_sana_sprint_img2img.py
@@ -84,7 +84,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -379,7 +379,7 @@ def encode_prompt(
 
         return prompt_embeds, prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/sana_video/pipeline_sana_video.py b/src/diffusers/pipelines/sana_video/pipeline_sana_video.py
index a786275e45a9..0a16977f8881 100644
--- a/src/diffusers/pipelines/sana_video/pipeline_sana_video.py
+++ b/src/diffusers/pipelines/sana_video/pipeline_sana_video.py
@@ -123,7 +123,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -419,7 +419,7 @@ def encode_prompt(
 
         return prompt_embeds, prompt_attention_mask, negative_prompt_embeds, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/sana_video/pipeline_sana_video_i2v.py b/src/diffusers/pipelines/sana_video/pipeline_sana_video_i2v.py
index e87880b64cee..e7b63cb0e4a6 100644
--- a/src/diffusers/pipelines/sana_video/pipeline_sana_video_i2v.py
+++ b/src/diffusers/pipelines/sana_video/pipeline_sana_video_i2v.py
@@ -99,7 +99,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -159,7 +159,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -417,7 +417,7 @@ def encode_prompt(
 
         return prompt_embeds, prompt_attention_mask, negative_prompt_embeds, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/semantic_stable_diffusion/pipeline_semantic_stable_diffusion.py b/src/diffusers/pipelines/semantic_stable_diffusion/pipeline_semantic_stable_diffusion.py
index 49b09e205cc5..0475df8e9b8f 100644
--- a/src/diffusers/pipelines/semantic_stable_diffusion/pipeline_semantic_stable_diffusion.py
+++ b/src/diffusers/pipelines/semantic_stable_diffusion/pipeline_semantic_stable_diffusion.py
@@ -99,7 +99,7 @@ def __init__(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -114,7 +114,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -126,7 +126,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/skyreels_v2/pipeline_skyreels_v2_diffusion_forcing.py b/src/diffusers/pipelines/skyreels_v2/pipeline_skyreels_v2_diffusion_forcing.py
index 089f92632d38..c610207ead01 100644
--- a/src/diffusers/pipelines/skyreels_v2/pipeline_skyreels_v2_diffusion_forcing.py
+++ b/src/diffusers/pipelines/skyreels_v2/pipeline_skyreels_v2_diffusion_forcing.py
@@ -112,7 +112,7 @@ def prompt_clean(text):
     return text
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/pipelines/skyreels_v2/pipeline_skyreels_v2_diffusion_forcing_i2v.py b/src/diffusers/pipelines/skyreels_v2/pipeline_skyreels_v2_diffusion_forcing_i2v.py
index 2951a9447386..e2a33f8d6cf7 100644
--- a/src/diffusers/pipelines/skyreels_v2/pipeline_skyreels_v2_diffusion_forcing_i2v.py
+++ b/src/diffusers/pipelines/skyreels_v2/pipeline_skyreels_v2_diffusion_forcing_i2v.py
@@ -117,7 +117,7 @@ def prompt_clean(text):
     return text
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/pipelines/skyreels_v2/pipeline_skyreels_v2_diffusion_forcing_v2v.py b/src/diffusers/pipelines/skyreels_v2/pipeline_skyreels_v2_diffusion_forcing_v2v.py
index 6fedfc795a40..e60252b69d29 100644
--- a/src/diffusers/pipelines/skyreels_v2/pipeline_skyreels_v2_diffusion_forcing_v2v.py
+++ b/src/diffusers/pipelines/skyreels_v2/pipeline_skyreels_v2_diffusion_forcing_v2v.py
@@ -113,7 +113,7 @@ def prompt_clean(text):
     return text
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -173,7 +173,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/pipelines/skyreels_v2/pipeline_skyreels_v2_i2v.py b/src/diffusers/pipelines/skyreels_v2/pipeline_skyreels_v2_i2v.py
index d61b687eadc3..78b706e24d9c 100644
--- a/src/diffusers/pipelines/skyreels_v2/pipeline_skyreels_v2_i2v.py
+++ b/src/diffusers/pipelines/skyreels_v2/pipeline_skyreels_v2_i2v.py
@@ -110,7 +110,7 @@ def prompt_clean(text):
     return text
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/pipelines/stable_audio/pipeline_stable_audio.py b/src/diffusers/pipelines/stable_audio/pipeline_stable_audio.py
index b7faf097ab0d..e3591883ac08 100644
--- a/src/diffusers/pipelines/stable_audio/pipeline_stable_audio.py
+++ b/src/diffusers/pipelines/stable_audio/pipeline_stable_audio.py
@@ -310,7 +310,7 @@ def encode_duration(
 
         return seconds_start_hidden_states, seconds_end_hidden_states
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py
index cb97f18efeff..e2745005b2ca 100644
--- a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py
+++ b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py
@@ -11,7 +11,6 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import torch
@@ -23,20 +22,12 @@
 from ...image_processor import PipelineImageInput, VaeImageProcessor
 from ...loaders import FromSingleFileMixin, IPAdapterMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, ImageProjection, UNet2DConditionModel
-from ...models.lora import adjust_lora_scale_text_encoder
 from ...schedulers import KarrasDiffusionSchedulers
-from ...utils import (
-    USE_PEFT_BACKEND,
-    deprecate,
-    is_torch_xla_available,
-    logging,
-    replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
-)
+from ...utils import deprecate, is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from .pipeline_output import StableDiffusionPipelineOutput
+from .pipeline_stable_diffusion_utils import SDMixin, rescale_noise_cfg, retrieve_timesteps
 from .safety_checker import StableDiffusionSafetyChecker
 
 
@@ -66,94 +57,10 @@
 """
 
 
-def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
-    r"""
-    Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
-    Section 3.4 from [Common Diffusion Noise Schedules and Sample Steps are
-    Flawed](https://huggingface.co/papers/2305.08891).
-
-    Args:
-        noise_cfg (`torch.Tensor`):
-            The predicted noise tensor for the guided diffusion process.
-        noise_pred_text (`torch.Tensor`):
-            The predicted noise tensor for the text-guided diffusion process.
-        guidance_rescale (`float`, *optional*, defaults to 0.0):
-            A rescale factor applied to the noise predictions.
-
-    Returns:
-        noise_cfg (`torch.Tensor`): The rescaled noise prediction tensor.
-    """
-    std_text = noise_pred_text.std(dim=list(range(1, noise_pred_text.ndim)), keepdim=True)
-    std_cfg = noise_cfg.std(dim=list(range(1, noise_cfg.ndim)), keepdim=True)
-    # rescale the results from guidance (fixes overexposure)
-    noise_pred_rescaled = noise_cfg * (std_text / std_cfg)
-    # mix with the original results from guidance by factor guidance_rescale to avoid "plain looking" images
-    noise_cfg = guidance_rescale * noise_pred_rescaled + (1 - guidance_rescale) * noise_cfg
-    return noise_cfg
-
-
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
 class StableDiffusionPipeline(
     DiffusionPipeline,
     StableDiffusionMixin,
+    SDMixin,
     TextualInversionLoaderMixin,
     StableDiffusionLoraLoaderMixin,
     IPAdapterMixin,
@@ -297,220 +204,6 @@ def __init__(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    def _encode_prompt(
-        self,
-        prompt,
-        device,
-        num_images_per_prompt,
-        do_classifier_free_guidance,
-        negative_prompt=None,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        negative_prompt_embeds: Optional[torch.Tensor] = None,
-        lora_scale: Optional[float] = None,
-        **kwargs,
-    ):
-        deprecation_message = "`_encode_prompt()` is deprecated and it will be removed in a future version. Use `encode_prompt()` instead. Also, be aware that the output format changed from a concatenated tensor to a tuple."
-        deprecate("_encode_prompt()", "1.0.0", deprecation_message, standard_warn=False)
-
-        prompt_embeds_tuple = self.encode_prompt(
-            prompt=prompt,
-            device=device,
-            num_images_per_prompt=num_images_per_prompt,
-            do_classifier_free_guidance=do_classifier_free_guidance,
-            negative_prompt=negative_prompt,
-            prompt_embeds=prompt_embeds,
-            negative_prompt_embeds=negative_prompt_embeds,
-            lora_scale=lora_scale,
-            **kwargs,
-        )
-
-        # concatenate for backwards comp
-        prompt_embeds = torch.cat([prompt_embeds_tuple[1], prompt_embeds_tuple[0]])
-
-        return prompt_embeds
-
-    def encode_prompt(
-        self,
-        prompt,
-        device,
-        num_images_per_prompt,
-        do_classifier_free_guidance,
-        negative_prompt=None,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        negative_prompt_embeds: Optional[torch.Tensor] = None,
-        lora_scale: Optional[float] = None,
-        clip_skip: Optional[int] = None,
-    ):
-        r"""
-        Encodes the prompt into text encoder hidden states.
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            do_classifier_free_guidance (`bool`):
-                whether to use classifier free guidance or not
-            negative_prompt (`str` or `List[str]`, *optional*):
-                The prompt or prompts not to guide the image generation. If not defined, one has to pass
-                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
-                less than `1`).
-            prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
-                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
-                argument.
-            lora_scale (`float`, *optional*):
-                A LoRA scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-            clip_skip (`int`, *optional*):
-                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
-                the output of the pre-final layer will be used for computing the prompt embeddings.
-        """
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if not USE_PEFT_BACKEND:
-                adjust_lora_scale_text_encoder(self.text_encoder, lora_scale)
-            else:
-                scale_lora_layers(self.text_encoder, lora_scale)
-
-        if prompt is not None and isinstance(prompt, str):
-            batch_size = 1
-        elif prompt is not None and isinstance(prompt, list):
-            batch_size = len(prompt)
-        else:
-            batch_size = prompt_embeds.shape[0]
-
-        if prompt_embeds is None:
-            # textual inversion: process multi-vector tokens if necessary
-            if isinstance(self, TextualInversionLoaderMixin):
-                prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-            text_inputs = self.tokenizer(
-                prompt,
-                padding="max_length",
-                max_length=self.tokenizer.model_max_length,
-                truncation=True,
-                return_tensors="pt",
-            )
-            text_input_ids = text_inputs.input_ids
-            untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-
-            if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
-                text_input_ids, untruncated_ids
-            ):
-                removed_text = self.tokenizer.batch_decode(
-                    untruncated_ids[:, self.tokenizer.model_max_length - 1 : -1]
-                )
-                logger.warning(
-                    "The following part of your input was truncated because CLIP can only handle sequences up to"
-                    f" {self.tokenizer.model_max_length} tokens: {removed_text}"
-                )
-
-            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
-                attention_mask = text_inputs.attention_mask.to(device)
-            else:
-                attention_mask = None
-
-            if clip_skip is None:
-                prompt_embeds = self.text_encoder(text_input_ids.to(device), attention_mask=attention_mask)
-                prompt_embeds = prompt_embeds[0]
-            else:
-                prompt_embeds = self.text_encoder(
-                    text_input_ids.to(device), attention_mask=attention_mask, output_hidden_states=True
-                )
-                # Access the `hidden_states` first, that contains a tuple of
-                # all the hidden states from the encoder layers. Then index into
-                # the tuple to access the hidden states from the desired layer.
-                prompt_embeds = prompt_embeds[-1][-(clip_skip + 1)]
-                # We also need to apply the final LayerNorm here to not mess with the
-                # representations. The `last_hidden_states` that we typically use for
-                # obtaining the final prompt representations passes through the LayerNorm
-                # layer.
-                prompt_embeds = self.text_encoder.text_model.final_layer_norm(prompt_embeds)
-
-        if self.text_encoder is not None:
-            prompt_embeds_dtype = self.text_encoder.dtype
-        elif self.unet is not None:
-            prompt_embeds_dtype = self.unet.dtype
-        else:
-            prompt_embeds_dtype = prompt_embeds.dtype
-
-        prompt_embeds = prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
-
-        bs_embed, seq_len, _ = prompt_embeds.shape
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)
-
-        # get unconditional embeddings for classifier free guidance
-        if do_classifier_free_guidance and negative_prompt_embeds is None:
-            uncond_tokens: List[str]
-            if negative_prompt is None:
-                uncond_tokens = [""] * batch_size
-            elif prompt is not None and type(prompt) is not type(negative_prompt):
-                raise TypeError(
-                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
-                    f" {type(prompt)}."
-                )
-            elif isinstance(negative_prompt, str):
-                uncond_tokens = [negative_prompt]
-            elif batch_size != len(negative_prompt):
-                raise ValueError(
-                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
-                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
-                    " the batch size of `prompt`."
-                )
-            else:
-                uncond_tokens = negative_prompt
-
-            # textual inversion: process multi-vector tokens if necessary
-            if isinstance(self, TextualInversionLoaderMixin):
-                uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)
-
-            max_length = prompt_embeds.shape[1]
-            uncond_input = self.tokenizer(
-                uncond_tokens,
-                padding="max_length",
-                max_length=max_length,
-                truncation=True,
-                return_tensors="pt",
-            )
-
-            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
-                attention_mask = uncond_input.attention_mask.to(device)
-            else:
-                attention_mask = None
-
-            negative_prompt_embeds = self.text_encoder(
-                uncond_input.input_ids.to(device),
-                attention_mask=attention_mask,
-            )
-            negative_prompt_embeds = negative_prompt_embeds[0]
-
-        if do_classifier_free_guidance:
-            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
-            seq_len = negative_prompt_embeds.shape[1]
-
-            negative_prompt_embeds = negative_prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
-
-            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
-            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        if self.text_encoder is not None:
-            if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        return prompt_embeds, negative_prompt_embeds
-
     def encode_image(self, image, device, num_images_per_prompt, output_hidden_states=None):
         dtype = next(self.image_encoder.parameters()).dtype
 
@@ -580,48 +273,6 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    def run_safety_checker(self, image, device, dtype):
-        if self.safety_checker is None:
-            has_nsfw_concept = None
-        else:
-            if torch.is_tensor(image):
-                feature_extractor_input = self.image_processor.postprocess(image, output_type="pil")
-            else:
-                feature_extractor_input = self.image_processor.numpy_to_pil(image)
-            safety_checker_input = self.feature_extractor(feature_extractor_input, return_tensors="pt").to(device)
-            image, has_nsfw_concept = self.safety_checker(
-                images=image, clip_input=safety_checker_input.pixel_values.to(dtype)
-            )
-        return image, has_nsfw_concept
-
-    def decode_latents(self, latents):
-        deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
-        deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
-
-        latents = 1 / self.vae.config.scaling_factor * latents
-        image = self.vae.decode(latents, return_dict=False)[0]
-        image = (image / 2 + 0.5).clamp(0, 1)
-        # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
-        image = image.cpu().permute(0, 2, 3, 1).float().numpy()
-        return image
-
-    def prepare_extra_step_kwargs(self, generator, eta):
-        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
-        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
-        # eta corresponds to η in DDIM paper: https://huggingface.co/papers/2010.02502
-        # and should be between [0, 1]
-
-        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        extra_step_kwargs = {}
-        if accepts_eta:
-            extra_step_kwargs["eta"] = eta
-
-        # check if the scheduler accepts generator
-        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        if accepts_generator:
-            extra_step_kwargs["generator"] = generator
-        return extra_step_kwargs
-
     def check_inputs(
         self,
         prompt,
@@ -744,37 +395,6 @@ def get_guidance_scale_embedding(
         assert emb.shape == (w.shape[0], embedding_dim)
         return emb
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def guidance_rescale(self):
-        return self._guidance_rescale
-
-    @property
-    def clip_skip(self):
-        return self._clip_skip
-
-    # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
-    # of the Imagen paper: https://huggingface.co/papers/2205.11487 . `guidance_scale = 1`
-    # corresponds to doing no classifier free guidance.
-    @property
-    def do_classifier_free_guidance(self):
-        return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
-
-    @property
-    def cross_attention_kwargs(self):
-        return self._cross_attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_depth2img.py b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_depth2img.py
index e957c6661f87..976c51be55d5 100644
--- a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_depth2img.py
+++ b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_depth2img.py
@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import contextlib
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -26,19 +25,16 @@
 from ...image_processor import PipelineImageInput, VaeImageProcessor
 from ...loaders import StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, UNet2DConditionModel
-from ...models.lora import adjust_lora_scale_text_encoder
 from ...schedulers import KarrasDiffusionSchedulers
 from ...utils import (
     PIL_INTERPOLATION,
-    USE_PEFT_BACKEND,
     deprecate,
     is_torch_xla_available,
     logging,
-    scale_lora_layers,
-    unscale_lora_layers,
 )
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput
+from .pipeline_stable_diffusion_utils import SDMixin, retrieve_latents
 
 
 if is_torch_xla_available():
@@ -51,20 +47,6 @@
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
 # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.preprocess
 def preprocess(image):
     deprecation_message = "The preprocess method is deprecated and will be removed in diffusers 1.0.0. Please use VaeImageProcessor.preprocess(...) instead"
@@ -89,7 +71,9 @@ def preprocess(image):
     return image
 
 
-class StableDiffusionDepth2ImgPipeline(DiffusionPipeline, TextualInversionLoaderMixin, StableDiffusionLoraLoaderMixin):
+class StableDiffusionDepth2ImgPipeline(
+    DiffusionPipeline, SDMixin, TextualInversionLoaderMixin, StableDiffusionLoraLoaderMixin
+):
     r"""
     Pipeline for text-guided depth-based image-to-image generation using Stable Diffusion.
 
@@ -167,267 +151,6 @@ def __init__(
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
-    def _encode_prompt(
-        self,
-        prompt,
-        device,
-        num_images_per_prompt,
-        do_classifier_free_guidance,
-        negative_prompt=None,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        negative_prompt_embeds: Optional[torch.Tensor] = None,
-        lora_scale: Optional[float] = None,
-        **kwargs,
-    ):
-        deprecation_message = "`_encode_prompt()` is deprecated and it will be removed in a future version. Use `encode_prompt()` instead. Also, be aware that the output format changed from a concatenated tensor to a tuple."
-        deprecate("_encode_prompt()", "1.0.0", deprecation_message, standard_warn=False)
-
-        prompt_embeds_tuple = self.encode_prompt(
-            prompt=prompt,
-            device=device,
-            num_images_per_prompt=num_images_per_prompt,
-            do_classifier_free_guidance=do_classifier_free_guidance,
-            negative_prompt=negative_prompt,
-            prompt_embeds=prompt_embeds,
-            negative_prompt_embeds=negative_prompt_embeds,
-            lora_scale=lora_scale,
-            **kwargs,
-        )
-
-        # concatenate for backwards comp
-        prompt_embeds = torch.cat([prompt_embeds_tuple[1], prompt_embeds_tuple[0]])
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt,
-        device,
-        num_images_per_prompt,
-        do_classifier_free_guidance,
-        negative_prompt=None,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        negative_prompt_embeds: Optional[torch.Tensor] = None,
-        lora_scale: Optional[float] = None,
-        clip_skip: Optional[int] = None,
-    ):
-        r"""
-        Encodes the prompt into text encoder hidden states.
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            do_classifier_free_guidance (`bool`):
-                whether to use classifier free guidance or not
-            negative_prompt (`str` or `List[str]`, *optional*):
-                The prompt or prompts not to guide the image generation. If not defined, one has to pass
-                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
-                less than `1`).
-            prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
-                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
-                argument.
-            lora_scale (`float`, *optional*):
-                A LoRA scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-            clip_skip (`int`, *optional*):
-                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
-                the output of the pre-final layer will be used for computing the prompt embeddings.
-        """
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if not USE_PEFT_BACKEND:
-                adjust_lora_scale_text_encoder(self.text_encoder, lora_scale)
-            else:
-                scale_lora_layers(self.text_encoder, lora_scale)
-
-        if prompt is not None and isinstance(prompt, str):
-            batch_size = 1
-        elif prompt is not None and isinstance(prompt, list):
-            batch_size = len(prompt)
-        else:
-            batch_size = prompt_embeds.shape[0]
-
-        if prompt_embeds is None:
-            # textual inversion: process multi-vector tokens if necessary
-            if isinstance(self, TextualInversionLoaderMixin):
-                prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-            text_inputs = self.tokenizer(
-                prompt,
-                padding="max_length",
-                max_length=self.tokenizer.model_max_length,
-                truncation=True,
-                return_tensors="pt",
-            )
-            text_input_ids = text_inputs.input_ids
-            untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-
-            if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
-                text_input_ids, untruncated_ids
-            ):
-                removed_text = self.tokenizer.batch_decode(
-                    untruncated_ids[:, self.tokenizer.model_max_length - 1 : -1]
-                )
-                logger.warning(
-                    "The following part of your input was truncated because CLIP can only handle sequences up to"
-                    f" {self.tokenizer.model_max_length} tokens: {removed_text}"
-                )
-
-            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
-                attention_mask = text_inputs.attention_mask.to(device)
-            else:
-                attention_mask = None
-
-            if clip_skip is None:
-                prompt_embeds = self.text_encoder(text_input_ids.to(device), attention_mask=attention_mask)
-                prompt_embeds = prompt_embeds[0]
-            else:
-                prompt_embeds = self.text_encoder(
-                    text_input_ids.to(device), attention_mask=attention_mask, output_hidden_states=True
-                )
-                # Access the `hidden_states` first, that contains a tuple of
-                # all the hidden states from the encoder layers. Then index into
-                # the tuple to access the hidden states from the desired layer.
-                prompt_embeds = prompt_embeds[-1][-(clip_skip + 1)]
-                # We also need to apply the final LayerNorm here to not mess with the
-                # representations. The `last_hidden_states` that we typically use for
-                # obtaining the final prompt representations passes through the LayerNorm
-                # layer.
-                prompt_embeds = self.text_encoder.text_model.final_layer_norm(prompt_embeds)
-
-        if self.text_encoder is not None:
-            prompt_embeds_dtype = self.text_encoder.dtype
-        elif self.unet is not None:
-            prompt_embeds_dtype = self.unet.dtype
-        else:
-            prompt_embeds_dtype = prompt_embeds.dtype
-
-        prompt_embeds = prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
-
-        bs_embed, seq_len, _ = prompt_embeds.shape
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)
-
-        # get unconditional embeddings for classifier free guidance
-        if do_classifier_free_guidance and negative_prompt_embeds is None:
-            uncond_tokens: List[str]
-            if negative_prompt is None:
-                uncond_tokens = [""] * batch_size
-            elif prompt is not None and type(prompt) is not type(negative_prompt):
-                raise TypeError(
-                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
-                    f" {type(prompt)}."
-                )
-            elif isinstance(negative_prompt, str):
-                uncond_tokens = [negative_prompt]
-            elif batch_size != len(negative_prompt):
-                raise ValueError(
-                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
-                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
-                    " the batch size of `prompt`."
-                )
-            else:
-                uncond_tokens = negative_prompt
-
-            # textual inversion: process multi-vector tokens if necessary
-            if isinstance(self, TextualInversionLoaderMixin):
-                uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)
-
-            max_length = prompt_embeds.shape[1]
-            uncond_input = self.tokenizer(
-                uncond_tokens,
-                padding="max_length",
-                max_length=max_length,
-                truncation=True,
-                return_tensors="pt",
-            )
-
-            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
-                attention_mask = uncond_input.attention_mask.to(device)
-            else:
-                attention_mask = None
-
-            negative_prompt_embeds = self.text_encoder(
-                uncond_input.input_ids.to(device),
-                attention_mask=attention_mask,
-            )
-            negative_prompt_embeds = negative_prompt_embeds[0]
-
-        if do_classifier_free_guidance:
-            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
-            seq_len = negative_prompt_embeds.shape[1]
-
-            negative_prompt_embeds = negative_prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
-
-            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
-            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        if self.text_encoder is not None:
-            if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        return prompt_embeds, negative_prompt_embeds
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
-    def run_safety_checker(self, image, device, dtype):
-        if self.safety_checker is None:
-            has_nsfw_concept = None
-        else:
-            if torch.is_tensor(image):
-                feature_extractor_input = self.image_processor.postprocess(image, output_type="pil")
-            else:
-                feature_extractor_input = self.image_processor.numpy_to_pil(image)
-            safety_checker_input = self.feature_extractor(feature_extractor_input, return_tensors="pt").to(device)
-            image, has_nsfw_concept = self.safety_checker(
-                images=image, clip_input=safety_checker_input.pixel_values.to(dtype)
-            )
-        return image, has_nsfw_concept
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
-    def decode_latents(self, latents):
-        deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
-        deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
-
-        latents = 1 / self.vae.config.scaling_factor * latents
-        image = self.vae.decode(latents, return_dict=False)[0]
-        image = (image / 2 + 0.5).clamp(0, 1)
-        # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
-        image = image.cpu().permute(0, 2, 3, 1).float().numpy()
-        return image
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
-    def prepare_extra_step_kwargs(self, generator, eta):
-        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
-        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
-        # eta corresponds to η in DDIM paper: https://huggingface.co/papers/2010.02502
-        # and should be between [0, 1]
-
-        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        extra_step_kwargs = {}
-        if accepts_eta:
-            extra_step_kwargs["eta"] = eta
-
-        # check if the scheduler accepts generator
-        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        if accepts_generator:
-            extra_step_kwargs["generator"] = generator
-        return extra_step_kwargs
-
     def check_inputs(
         self,
         prompt,
@@ -608,29 +331,6 @@ def prepare_depth_map(self, image, depth_map, batch_size, do_classifier_free_gui
         depth_map = torch.cat([depth_map] * 2) if do_classifier_free_guidance else depth_map
         return depth_map
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def clip_skip(self):
-        return self._clip_skip
-
-    # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
-    # of the Imagen paper: https://huggingface.co/papers/2205.11487 . `guidance_scale = 1`
-    # corresponds to doing no classifier free guidance.
-    @property
-    def do_classifier_free_guidance(self):
-        return self._guidance_scale > 1
-
-    @property
-    def cross_attention_kwargs(self):
-        return self._cross_attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
     @torch.no_grad()
     def __call__(
         self,
diff --git a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_image_variation.py b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_image_variation.py
index d47e2f0593dd..d9e695312b4d 100644
--- a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_image_variation.py
+++ b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_image_variation.py
@@ -166,7 +166,7 @@ def _encode_image(self, image, device, num_images_per_prompt, do_classifier_free
 
         return image_embeddings
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -181,7 +181,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -193,7 +193,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py
index 95d3ab06f02a..b6aa1c25227a 100644
--- a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py
+++ b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import numpy as np
@@ -26,21 +25,18 @@
 from ...image_processor import PipelineImageInput, VaeImageProcessor
 from ...loaders import FromSingleFileMixin, IPAdapterMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, ImageProjection, UNet2DConditionModel
-from ...models.lora import adjust_lora_scale_text_encoder
 from ...schedulers import KarrasDiffusionSchedulers
 from ...utils import (
     PIL_INTERPOLATION,
-    USE_PEFT_BACKEND,
     deprecate,
     is_torch_xla_available,
     logging,
     replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
 )
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from . import StableDiffusionPipelineOutput
+from .pipeline_stable_diffusion_utils import SDMixin, retrieve_latents, retrieve_timesteps
 from .safety_checker import StableDiffusionSafetyChecker
 
 
@@ -83,19 +79,6 @@
 """
 
 
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
 def preprocess(image):
     deprecation_message = "The preprocess method is deprecated and will be removed in diffusers 1.0.0. Please use VaeImageProcessor.preprocess(...) instead"
     deprecate("preprocess", "1.0.0", deprecation_message, standard_warn=False)
@@ -119,69 +102,10 @@ def preprocess(image):
     return image
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
 class StableDiffusionImg2ImgPipeline(
     DiffusionPipeline,
     StableDiffusionMixin,
+    SDMixin,
     TextualInversionLoaderMixin,
     IPAdapterMixin,
     StableDiffusionLoraLoaderMixin,
@@ -321,222 +245,6 @@ def __init__(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
-    def _encode_prompt(
-        self,
-        prompt,
-        device,
-        num_images_per_prompt,
-        do_classifier_free_guidance,
-        negative_prompt=None,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        negative_prompt_embeds: Optional[torch.Tensor] = None,
-        lora_scale: Optional[float] = None,
-        **kwargs,
-    ):
-        deprecation_message = "`_encode_prompt()` is deprecated and it will be removed in a future version. Use `encode_prompt()` instead. Also, be aware that the output format changed from a concatenated tensor to a tuple."
-        deprecate("_encode_prompt()", "1.0.0", deprecation_message, standard_warn=False)
-
-        prompt_embeds_tuple = self.encode_prompt(
-            prompt=prompt,
-            device=device,
-            num_images_per_prompt=num_images_per_prompt,
-            do_classifier_free_guidance=do_classifier_free_guidance,
-            negative_prompt=negative_prompt,
-            prompt_embeds=prompt_embeds,
-            negative_prompt_embeds=negative_prompt_embeds,
-            lora_scale=lora_scale,
-            **kwargs,
-        )
-
-        # concatenate for backwards comp
-        prompt_embeds = torch.cat([prompt_embeds_tuple[1], prompt_embeds_tuple[0]])
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt,
-        device,
-        num_images_per_prompt,
-        do_classifier_free_guidance,
-        negative_prompt=None,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        negative_prompt_embeds: Optional[torch.Tensor] = None,
-        lora_scale: Optional[float] = None,
-        clip_skip: Optional[int] = None,
-    ):
-        r"""
-        Encodes the prompt into text encoder hidden states.
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            do_classifier_free_guidance (`bool`):
-                whether to use classifier free guidance or not
-            negative_prompt (`str` or `List[str]`, *optional*):
-                The prompt or prompts not to guide the image generation. If not defined, one has to pass
-                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
-                less than `1`).
-            prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
-                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
-                argument.
-            lora_scale (`float`, *optional*):
-                A LoRA scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-            clip_skip (`int`, *optional*):
-                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
-                the output of the pre-final layer will be used for computing the prompt embeddings.
-        """
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if not USE_PEFT_BACKEND:
-                adjust_lora_scale_text_encoder(self.text_encoder, lora_scale)
-            else:
-                scale_lora_layers(self.text_encoder, lora_scale)
-
-        if prompt is not None and isinstance(prompt, str):
-            batch_size = 1
-        elif prompt is not None and isinstance(prompt, list):
-            batch_size = len(prompt)
-        else:
-            batch_size = prompt_embeds.shape[0]
-
-        if prompt_embeds is None:
-            # textual inversion: process multi-vector tokens if necessary
-            if isinstance(self, TextualInversionLoaderMixin):
-                prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-            text_inputs = self.tokenizer(
-                prompt,
-                padding="max_length",
-                max_length=self.tokenizer.model_max_length,
-                truncation=True,
-                return_tensors="pt",
-            )
-            text_input_ids = text_inputs.input_ids
-            untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-
-            if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
-                text_input_ids, untruncated_ids
-            ):
-                removed_text = self.tokenizer.batch_decode(
-                    untruncated_ids[:, self.tokenizer.model_max_length - 1 : -1]
-                )
-                logger.warning(
-                    "The following part of your input was truncated because CLIP can only handle sequences up to"
-                    f" {self.tokenizer.model_max_length} tokens: {removed_text}"
-                )
-
-            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
-                attention_mask = text_inputs.attention_mask.to(device)
-            else:
-                attention_mask = None
-
-            if clip_skip is None:
-                prompt_embeds = self.text_encoder(text_input_ids.to(device), attention_mask=attention_mask)
-                prompt_embeds = prompt_embeds[0]
-            else:
-                prompt_embeds = self.text_encoder(
-                    text_input_ids.to(device), attention_mask=attention_mask, output_hidden_states=True
-                )
-                # Access the `hidden_states` first, that contains a tuple of
-                # all the hidden states from the encoder layers. Then index into
-                # the tuple to access the hidden states from the desired layer.
-                prompt_embeds = prompt_embeds[-1][-(clip_skip + 1)]
-                # We also need to apply the final LayerNorm here to not mess with the
-                # representations. The `last_hidden_states` that we typically use for
-                # obtaining the final prompt representations passes through the LayerNorm
-                # layer.
-                prompt_embeds = self.text_encoder.text_model.final_layer_norm(prompt_embeds)
-
-        if self.text_encoder is not None:
-            prompt_embeds_dtype = self.text_encoder.dtype
-        elif self.unet is not None:
-            prompt_embeds_dtype = self.unet.dtype
-        else:
-            prompt_embeds_dtype = prompt_embeds.dtype
-
-        prompt_embeds = prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
-
-        bs_embed, seq_len, _ = prompt_embeds.shape
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)
-
-        # get unconditional embeddings for classifier free guidance
-        if do_classifier_free_guidance and negative_prompt_embeds is None:
-            uncond_tokens: List[str]
-            if negative_prompt is None:
-                uncond_tokens = [""] * batch_size
-            elif prompt is not None and type(prompt) is not type(negative_prompt):
-                raise TypeError(
-                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
-                    f" {type(prompt)}."
-                )
-            elif isinstance(negative_prompt, str):
-                uncond_tokens = [negative_prompt]
-            elif batch_size != len(negative_prompt):
-                raise ValueError(
-                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
-                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
-                    " the batch size of `prompt`."
-                )
-            else:
-                uncond_tokens = negative_prompt
-
-            # textual inversion: process multi-vector tokens if necessary
-            if isinstance(self, TextualInversionLoaderMixin):
-                uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)
-
-            max_length = prompt_embeds.shape[1]
-            uncond_input = self.tokenizer(
-                uncond_tokens,
-                padding="max_length",
-                max_length=max_length,
-                truncation=True,
-                return_tensors="pt",
-            )
-
-            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
-                attention_mask = uncond_input.attention_mask.to(device)
-            else:
-                attention_mask = None
-
-            negative_prompt_embeds = self.text_encoder(
-                uncond_input.input_ids.to(device),
-                attention_mask=attention_mask,
-            )
-            negative_prompt_embeds = negative_prompt_embeds[0]
-
-        if do_classifier_free_guidance:
-            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
-            seq_len = negative_prompt_embeds.shape[1]
-
-            negative_prompt_embeds = negative_prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
-
-            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
-            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        if self.text_encoder is not None:
-            if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        return prompt_embeds, negative_prompt_embeds
-
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_image
     def encode_image(self, image, device, num_images_per_prompt, output_hidden_states=None):
         dtype = next(self.image_encoder.parameters()).dtype
@@ -608,51 +316,6 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
-    def run_safety_checker(self, image, device, dtype):
-        if self.safety_checker is None:
-            has_nsfw_concept = None
-        else:
-            if torch.is_tensor(image):
-                feature_extractor_input = self.image_processor.postprocess(image, output_type="pil")
-            else:
-                feature_extractor_input = self.image_processor.numpy_to_pil(image)
-            safety_checker_input = self.feature_extractor(feature_extractor_input, return_tensors="pt").to(device)
-            image, has_nsfw_concept = self.safety_checker(
-                images=image, clip_input=safety_checker_input.pixel_values.to(dtype)
-            )
-        return image, has_nsfw_concept
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
-    def decode_latents(self, latents):
-        deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
-        deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
-
-        latents = 1 / self.vae.config.scaling_factor * latents
-        image = self.vae.decode(latents, return_dict=False)[0]
-        image = (image / 2 + 0.5).clamp(0, 1)
-        # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
-        image = image.cpu().permute(0, 2, 3, 1).float().numpy()
-        return image
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
-    def prepare_extra_step_kwargs(self, generator, eta):
-        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
-        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
-        # eta corresponds to η in DDIM paper: https://huggingface.co/papers/2010.02502
-        # and should be between [0, 1]
-
-        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        extra_step_kwargs = {}
-        if accepts_eta:
-            extra_step_kwargs["eta"] = eta
-
-        # check if the scheduler accepts generator
-        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        if accepts_generator:
-            extra_step_kwargs["generator"] = generator
-        return extra_step_kwargs
-
     def check_inputs(
         self,
         prompt,
@@ -828,33 +491,6 @@ def get_guidance_scale_embedding(
         assert emb.shape == (w.shape[0], embedding_dim)
         return emb
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def clip_skip(self):
-        return self._clip_skip
-
-    # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
-    # of the Imagen paper: https://huggingface.co/papers/2205.11487 . `guidance_scale = 1`
-    # corresponds to doing no classifier free guidance.
-    @property
-    def do_classifier_free_guidance(self):
-        return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
-
-    @property
-    def cross_attention_kwargs(self):
-        return self._cross_attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py
index 148d7386a732..3595f837f92b 100644
--- a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py
+++ b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import PIL.Image
@@ -25,19 +24,12 @@
 from ...image_processor import PipelineImageInput, VaeImageProcessor
 from ...loaders import FromSingleFileMixin, IPAdapterMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AsymmetricAutoencoderKL, AutoencoderKL, ImageProjection, UNet2DConditionModel
-from ...models.lora import adjust_lora_scale_text_encoder
 from ...schedulers import KarrasDiffusionSchedulers
-from ...utils import (
-    USE_PEFT_BACKEND,
-    deprecate,
-    is_torch_xla_available,
-    logging,
-    scale_lora_layers,
-    unscale_lora_layers,
-)
+from ...utils import deprecate, is_torch_xla_available, logging
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from . import StableDiffusionPipelineOutput
+from .pipeline_stable_diffusion_utils import SDMixin, retrieve_latents, retrieve_timesteps
 from .safety_checker import StableDiffusionSafetyChecker
 
 
@@ -51,83 +43,10 @@
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
 class StableDiffusionInpaintPipeline(
     DiffusionPipeline,
     StableDiffusionMixin,
+    SDMixin,
     TextualInversionLoaderMixin,
     IPAdapterMixin,
     StableDiffusionLoraLoaderMixin,
@@ -275,222 +194,6 @@ def __init__(
         )
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
-    def _encode_prompt(
-        self,
-        prompt,
-        device,
-        num_images_per_prompt,
-        do_classifier_free_guidance,
-        negative_prompt=None,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        negative_prompt_embeds: Optional[torch.Tensor] = None,
-        lora_scale: Optional[float] = None,
-        **kwargs,
-    ):
-        deprecation_message = "`_encode_prompt()` is deprecated and it will be removed in a future version. Use `encode_prompt()` instead. Also, be aware that the output format changed from a concatenated tensor to a tuple."
-        deprecate("_encode_prompt()", "1.0.0", deprecation_message, standard_warn=False)
-
-        prompt_embeds_tuple = self.encode_prompt(
-            prompt=prompt,
-            device=device,
-            num_images_per_prompt=num_images_per_prompt,
-            do_classifier_free_guidance=do_classifier_free_guidance,
-            negative_prompt=negative_prompt,
-            prompt_embeds=prompt_embeds,
-            negative_prompt_embeds=negative_prompt_embeds,
-            lora_scale=lora_scale,
-            **kwargs,
-        )
-
-        # concatenate for backwards comp
-        prompt_embeds = torch.cat([prompt_embeds_tuple[1], prompt_embeds_tuple[0]])
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt,
-        device,
-        num_images_per_prompt,
-        do_classifier_free_guidance,
-        negative_prompt=None,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        negative_prompt_embeds: Optional[torch.Tensor] = None,
-        lora_scale: Optional[float] = None,
-        clip_skip: Optional[int] = None,
-    ):
-        r"""
-        Encodes the prompt into text encoder hidden states.
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            do_classifier_free_guidance (`bool`):
-                whether to use classifier free guidance or not
-            negative_prompt (`str` or `List[str]`, *optional*):
-                The prompt or prompts not to guide the image generation. If not defined, one has to pass
-                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
-                less than `1`).
-            prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
-                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
-                argument.
-            lora_scale (`float`, *optional*):
-                A LoRA scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-            clip_skip (`int`, *optional*):
-                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
-                the output of the pre-final layer will be used for computing the prompt embeddings.
-        """
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if not USE_PEFT_BACKEND:
-                adjust_lora_scale_text_encoder(self.text_encoder, lora_scale)
-            else:
-                scale_lora_layers(self.text_encoder, lora_scale)
-
-        if prompt is not None and isinstance(prompt, str):
-            batch_size = 1
-        elif prompt is not None and isinstance(prompt, list):
-            batch_size = len(prompt)
-        else:
-            batch_size = prompt_embeds.shape[0]
-
-        if prompt_embeds is None:
-            # textual inversion: process multi-vector tokens if necessary
-            if isinstance(self, TextualInversionLoaderMixin):
-                prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-            text_inputs = self.tokenizer(
-                prompt,
-                padding="max_length",
-                max_length=self.tokenizer.model_max_length,
-                truncation=True,
-                return_tensors="pt",
-            )
-            text_input_ids = text_inputs.input_ids
-            untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-
-            if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
-                text_input_ids, untruncated_ids
-            ):
-                removed_text = self.tokenizer.batch_decode(
-                    untruncated_ids[:, self.tokenizer.model_max_length - 1 : -1]
-                )
-                logger.warning(
-                    "The following part of your input was truncated because CLIP can only handle sequences up to"
-                    f" {self.tokenizer.model_max_length} tokens: {removed_text}"
-                )
-
-            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
-                attention_mask = text_inputs.attention_mask.to(device)
-            else:
-                attention_mask = None
-
-            if clip_skip is None:
-                prompt_embeds = self.text_encoder(text_input_ids.to(device), attention_mask=attention_mask)
-                prompt_embeds = prompt_embeds[0]
-            else:
-                prompt_embeds = self.text_encoder(
-                    text_input_ids.to(device), attention_mask=attention_mask, output_hidden_states=True
-                )
-                # Access the `hidden_states` first, that contains a tuple of
-                # all the hidden states from the encoder layers. Then index into
-                # the tuple to access the hidden states from the desired layer.
-                prompt_embeds = prompt_embeds[-1][-(clip_skip + 1)]
-                # We also need to apply the final LayerNorm here to not mess with the
-                # representations. The `last_hidden_states` that we typically use for
-                # obtaining the final prompt representations passes through the LayerNorm
-                # layer.
-                prompt_embeds = self.text_encoder.text_model.final_layer_norm(prompt_embeds)
-
-        if self.text_encoder is not None:
-            prompt_embeds_dtype = self.text_encoder.dtype
-        elif self.unet is not None:
-            prompt_embeds_dtype = self.unet.dtype
-        else:
-            prompt_embeds_dtype = prompt_embeds.dtype
-
-        prompt_embeds = prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
-
-        bs_embed, seq_len, _ = prompt_embeds.shape
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)
-
-        # get unconditional embeddings for classifier free guidance
-        if do_classifier_free_guidance and negative_prompt_embeds is None:
-            uncond_tokens: List[str]
-            if negative_prompt is None:
-                uncond_tokens = [""] * batch_size
-            elif prompt is not None and type(prompt) is not type(negative_prompt):
-                raise TypeError(
-                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
-                    f" {type(prompt)}."
-                )
-            elif isinstance(negative_prompt, str):
-                uncond_tokens = [negative_prompt]
-            elif batch_size != len(negative_prompt):
-                raise ValueError(
-                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
-                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
-                    " the batch size of `prompt`."
-                )
-            else:
-                uncond_tokens = negative_prompt
-
-            # textual inversion: process multi-vector tokens if necessary
-            if isinstance(self, TextualInversionLoaderMixin):
-                uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)
-
-            max_length = prompt_embeds.shape[1]
-            uncond_input = self.tokenizer(
-                uncond_tokens,
-                padding="max_length",
-                max_length=max_length,
-                truncation=True,
-                return_tensors="pt",
-            )
-
-            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
-                attention_mask = uncond_input.attention_mask.to(device)
-            else:
-                attention_mask = None
-
-            negative_prompt_embeds = self.text_encoder(
-                uncond_input.input_ids.to(device),
-                attention_mask=attention_mask,
-            )
-            negative_prompt_embeds = negative_prompt_embeds[0]
-
-        if do_classifier_free_guidance:
-            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
-            seq_len = negative_prompt_embeds.shape[1]
-
-            negative_prompt_embeds = negative_prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
-
-            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
-            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        if self.text_encoder is not None:
-            if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        return prompt_embeds, negative_prompt_embeds
-
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_image
     def encode_image(self, image, device, num_images_per_prompt, output_hidden_states=None):
         dtype = next(self.image_encoder.parameters()).dtype
@@ -562,39 +265,6 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
-    def run_safety_checker(self, image, device, dtype):
-        if self.safety_checker is None:
-            has_nsfw_concept = None
-        else:
-            if torch.is_tensor(image):
-                feature_extractor_input = self.image_processor.postprocess(image, output_type="pil")
-            else:
-                feature_extractor_input = self.image_processor.numpy_to_pil(image)
-            safety_checker_input = self.feature_extractor(feature_extractor_input, return_tensors="pt").to(device)
-            image, has_nsfw_concept = self.safety_checker(
-                images=image, clip_input=safety_checker_input.pixel_values.to(dtype)
-            )
-        return image, has_nsfw_concept
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
-    def prepare_extra_step_kwargs(self, generator, eta):
-        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
-        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
-        # eta corresponds to η in DDIM paper: https://huggingface.co/papers/2010.02502
-        # and should be between [0, 1]
-
-        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        extra_step_kwargs = {}
-        if accepts_eta:
-            extra_step_kwargs["eta"] = eta
-
-        # check if the scheduler accepts generator
-        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        if accepts_generator:
-            extra_step_kwargs["generator"] = generator
-        return extra_step_kwargs
-
     def check_inputs(
         self,
         prompt,
@@ -850,33 +520,6 @@ def get_guidance_scale_embedding(
         assert emb.shape == (w.shape[0], embedding_dim)
         return emb
 
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def clip_skip(self):
-        return self._clip_skip
-
-    # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
-    # of the Imagen paper: https://huggingface.co/papers/2205.11487 . `guidance_scale = 1`
-    # corresponds to doing no classifier free guidance.
-    @property
-    def do_classifier_free_guidance(self):
-        return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
-
-    @property
-    def cross_attention_kwargs(self):
-        return self._cross_attention_kwargs
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     def __call__(
         self,
diff --git a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_instruct_pix2pix.py b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_instruct_pix2pix.py
index 843d25d67c10..a6c98e9f2580 100644
--- a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_instruct_pix2pix.py
+++ b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_instruct_pix2pix.py
@@ -66,7 +66,7 @@ def preprocess(image):
     return image
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -723,7 +723,7 @@ def prepare_ip_adapter_image_embeds(
 
         return image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -738,7 +738,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -756,7 +756,7 @@ def prepare_extra_step_kwargs(self, generator, eta):
             extra_step_kwargs["generator"] = generator
         return extra_step_kwargs
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
diff --git a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_latent_upscale.py b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_latent_upscale.py
index 66d5ffa6b849..66f27235a05c 100644
--- a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_latent_upscale.py
+++ b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_latent_upscale.py
@@ -28,6 +28,7 @@
 from ...utils import deprecate, is_torch_xla_available, logging
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput, StableDiffusionMixin
+from .pipeline_stable_diffusion_utils import SDMixin, retrieve_latents
 
 
 if is_torch_xla_available():
@@ -40,20 +41,6 @@
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
 # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_upscale.preprocess
 def preprocess(image):
     warnings.warn(
@@ -81,7 +68,7 @@ def preprocess(image):
     return image
 
 
-class StableDiffusionLatentUpscalePipeline(DiffusionPipeline, StableDiffusionMixin, FromSingleFileMixin):
+class StableDiffusionLatentUpscalePipeline(DiffusionPipeline, StableDiffusionMixin, SDMixin, FromSingleFileMixin):
     r"""
     Pipeline for upscaling Stable Diffusion output image resolution by a factor of 2.
 
@@ -282,18 +269,6 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
-    def decode_latents(self, latents):
-        deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
-        deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
-
-        latents = 1 / self.vae.config.scaling_factor * latents
-        image = self.vae.decode(latents, return_dict=False)[0]
-        image = (image / 2 + 0.5).clamp(0, 1)
-        # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
-        image = image.cpu().permute(0, 2, 3, 1).float().numpy()
-        return image
-
     def check_inputs(
         self,
         prompt,
diff --git a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_upscale.py b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_upscale.py
index 02f7526b7102..a6c94745fbeb 100644
--- a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_upscale.py
+++ b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_upscale.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 import warnings
 from typing import Any, Callable, Dict, List, Optional, Union
 
@@ -24,23 +23,13 @@
 from ...image_processor import PipelineImageInput, VaeImageProcessor
 from ...loaders import FromSingleFileMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, UNet2DConditionModel
-from ...models.attention_processor import (
-    AttnProcessor2_0,
-    XFormersAttnProcessor,
-)
-from ...models.lora import adjust_lora_scale_text_encoder
+from ...models.attention_processor import AttnProcessor2_0, XFormersAttnProcessor
 from ...schedulers import DDPMScheduler, KarrasDiffusionSchedulers
-from ...utils import (
-    USE_PEFT_BACKEND,
-    deprecate,
-    is_torch_xla_available,
-    logging,
-    scale_lora_layers,
-    unscale_lora_layers,
-)
+from ...utils import deprecate, is_torch_xla_available, logging
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from . import StableDiffusionPipelineOutput
+from .pipeline_stable_diffusion_utils import SDMixin
 
 
 if is_torch_xla_available():
@@ -82,6 +71,7 @@ def preprocess(image):
 class StableDiffusionUpscalePipeline(
     DiffusionPipeline,
     StableDiffusionMixin,
+    SDMixin,
     TextualInversionLoaderMixin,
     StableDiffusionLoraLoaderMixin,
     FromSingleFileMixin,
@@ -184,252 +174,6 @@ def run_safety_checker(self, image, device, dtype):
 
         return image, nsfw_detected, watermark_detected
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
-    def _encode_prompt(
-        self,
-        prompt,
-        device,
-        num_images_per_prompt,
-        do_classifier_free_guidance,
-        negative_prompt=None,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        negative_prompt_embeds: Optional[torch.Tensor] = None,
-        lora_scale: Optional[float] = None,
-        **kwargs,
-    ):
-        deprecation_message = "`_encode_prompt()` is deprecated and it will be removed in a future version. Use `encode_prompt()` instead. Also, be aware that the output format changed from a concatenated tensor to a tuple."
-        deprecate("_encode_prompt()", "1.0.0", deprecation_message, standard_warn=False)
-
-        prompt_embeds_tuple = self.encode_prompt(
-            prompt=prompt,
-            device=device,
-            num_images_per_prompt=num_images_per_prompt,
-            do_classifier_free_guidance=do_classifier_free_guidance,
-            negative_prompt=negative_prompt,
-            prompt_embeds=prompt_embeds,
-            negative_prompt_embeds=negative_prompt_embeds,
-            lora_scale=lora_scale,
-            **kwargs,
-        )
-
-        # concatenate for backwards comp
-        prompt_embeds = torch.cat([prompt_embeds_tuple[1], prompt_embeds_tuple[0]])
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt,
-        device,
-        num_images_per_prompt,
-        do_classifier_free_guidance,
-        negative_prompt=None,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        negative_prompt_embeds: Optional[torch.Tensor] = None,
-        lora_scale: Optional[float] = None,
-        clip_skip: Optional[int] = None,
-    ):
-        r"""
-        Encodes the prompt into text encoder hidden states.
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            do_classifier_free_guidance (`bool`):
-                whether to use classifier free guidance or not
-            negative_prompt (`str` or `List[str]`, *optional*):
-                The prompt or prompts not to guide the image generation. If not defined, one has to pass
-                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
-                less than `1`).
-            prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
-                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
-                argument.
-            lora_scale (`float`, *optional*):
-                A LoRA scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-            clip_skip (`int`, *optional*):
-                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
-                the output of the pre-final layer will be used for computing the prompt embeddings.
-        """
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if not USE_PEFT_BACKEND:
-                adjust_lora_scale_text_encoder(self.text_encoder, lora_scale)
-            else:
-                scale_lora_layers(self.text_encoder, lora_scale)
-
-        if prompt is not None and isinstance(prompt, str):
-            batch_size = 1
-        elif prompt is not None and isinstance(prompt, list):
-            batch_size = len(prompt)
-        else:
-            batch_size = prompt_embeds.shape[0]
-
-        if prompt_embeds is None:
-            # textual inversion: process multi-vector tokens if necessary
-            if isinstance(self, TextualInversionLoaderMixin):
-                prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-            text_inputs = self.tokenizer(
-                prompt,
-                padding="max_length",
-                max_length=self.tokenizer.model_max_length,
-                truncation=True,
-                return_tensors="pt",
-            )
-            text_input_ids = text_inputs.input_ids
-            untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-
-            if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
-                text_input_ids, untruncated_ids
-            ):
-                removed_text = self.tokenizer.batch_decode(
-                    untruncated_ids[:, self.tokenizer.model_max_length - 1 : -1]
-                )
-                logger.warning(
-                    "The following part of your input was truncated because CLIP can only handle sequences up to"
-                    f" {self.tokenizer.model_max_length} tokens: {removed_text}"
-                )
-
-            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
-                attention_mask = text_inputs.attention_mask.to(device)
-            else:
-                attention_mask = None
-
-            if clip_skip is None:
-                prompt_embeds = self.text_encoder(text_input_ids.to(device), attention_mask=attention_mask)
-                prompt_embeds = prompt_embeds[0]
-            else:
-                prompt_embeds = self.text_encoder(
-                    text_input_ids.to(device), attention_mask=attention_mask, output_hidden_states=True
-                )
-                # Access the `hidden_states` first, that contains a tuple of
-                # all the hidden states from the encoder layers. Then index into
-                # the tuple to access the hidden states from the desired layer.
-                prompt_embeds = prompt_embeds[-1][-(clip_skip + 1)]
-                # We also need to apply the final LayerNorm here to not mess with the
-                # representations. The `last_hidden_states` that we typically use for
-                # obtaining the final prompt representations passes through the LayerNorm
-                # layer.
-                prompt_embeds = self.text_encoder.text_model.final_layer_norm(prompt_embeds)
-
-        if self.text_encoder is not None:
-            prompt_embeds_dtype = self.text_encoder.dtype
-        elif self.unet is not None:
-            prompt_embeds_dtype = self.unet.dtype
-        else:
-            prompt_embeds_dtype = prompt_embeds.dtype
-
-        prompt_embeds = prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
-
-        bs_embed, seq_len, _ = prompt_embeds.shape
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)
-
-        # get unconditional embeddings for classifier free guidance
-        if do_classifier_free_guidance and negative_prompt_embeds is None:
-            uncond_tokens: List[str]
-            if negative_prompt is None:
-                uncond_tokens = [""] * batch_size
-            elif prompt is not None and type(prompt) is not type(negative_prompt):
-                raise TypeError(
-                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
-                    f" {type(prompt)}."
-                )
-            elif isinstance(negative_prompt, str):
-                uncond_tokens = [negative_prompt]
-            elif batch_size != len(negative_prompt):
-                raise ValueError(
-                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
-                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
-                    " the batch size of `prompt`."
-                )
-            else:
-                uncond_tokens = negative_prompt
-
-            # textual inversion: process multi-vector tokens if necessary
-            if isinstance(self, TextualInversionLoaderMixin):
-                uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)
-
-            max_length = prompt_embeds.shape[1]
-            uncond_input = self.tokenizer(
-                uncond_tokens,
-                padding="max_length",
-                max_length=max_length,
-                truncation=True,
-                return_tensors="pt",
-            )
-
-            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
-                attention_mask = uncond_input.attention_mask.to(device)
-            else:
-                attention_mask = None
-
-            negative_prompt_embeds = self.text_encoder(
-                uncond_input.input_ids.to(device),
-                attention_mask=attention_mask,
-            )
-            negative_prompt_embeds = negative_prompt_embeds[0]
-
-        if do_classifier_free_guidance:
-            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
-            seq_len = negative_prompt_embeds.shape[1]
-
-            negative_prompt_embeds = negative_prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
-
-            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
-            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        if self.text_encoder is not None:
-            if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        return prompt_embeds, negative_prompt_embeds
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
-    def prepare_extra_step_kwargs(self, generator, eta):
-        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
-        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
-        # eta corresponds to η in DDIM paper: https://huggingface.co/papers/2010.02502
-        # and should be between [0, 1]
-
-        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        extra_step_kwargs = {}
-        if accepts_eta:
-            extra_step_kwargs["eta"] = eta
-
-        # check if the scheduler accepts generator
-        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        if accepts_generator:
-            extra_step_kwargs["generator"] = generator
-        return extra_step_kwargs
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
-    def decode_latents(self, latents):
-        deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
-        deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
-
-        latents = 1 / self.vae.config.scaling_factor * latents
-        image = self.vae.decode(latents, return_dict=False)[0]
-        image = (image / 2 + 0.5).clamp(0, 1)
-        # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
-        image = image.cpu().permute(0, 2, 3, 1).float().numpy()
-        return image
-
     def check_inputs(
         self,
         prompt,
@@ -533,10 +277,7 @@ def upcast_vae(self):
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_utils.py b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_utils.py
new file mode 100644
index 000000000000..7354effc0eba
--- /dev/null
+++ b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_utils.py
@@ -0,0 +1,398 @@
+import inspect
+from typing import List, Optional, Union
+
+import torch
+
+from ...loaders import StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
+from ...models.lora import adjust_lora_scale_text_encoder
+from ...utils import USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers, unscale_lora_layers
+
+
+logger = logging.get_logger(__name__)
+
+
+def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
+    r"""
+    Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
+    Section 3.4 from [Common Diffusion Noise Schedules and Sample Steps are
+    Flawed](https://huggingface.co/papers/2305.08891).
+
+    Args:
+        noise_cfg (`torch.Tensor`):
+            The predicted noise tensor for the guided diffusion process.
+        noise_pred_text (`torch.Tensor`):
+            The predicted noise tensor for the text-guided diffusion process.
+        guidance_rescale (`float`, *optional*, defaults to 0.0):
+            A rescale factor applied to the noise predictions.
+
+    Returns:
+        noise_cfg (`torch.Tensor`): The rescaled noise prediction tensor.
+    """
+    std_text = noise_pred_text.std(dim=list(range(1, noise_pred_text.ndim)), keepdim=True)
+    std_cfg = noise_cfg.std(dim=list(range(1, noise_cfg.ndim)), keepdim=True)
+    # rescale the results from guidance (fixes overexposure)
+    noise_pred_rescaled = noise_cfg * (std_text / std_cfg)
+    # mix with the original results from guidance by factor guidance_rescale to avoid "plain looking" images
+    noise_cfg = guidance_rescale * noise_pred_rescaled + (1 - guidance_rescale) * noise_cfg
+    return noise_cfg
+
+
+def retrieve_timesteps(
+    scheduler,
+    num_inference_steps: Optional[int] = None,
+    device: Optional[Union[str, torch.device]] = None,
+    timesteps: Optional[List[int]] = None,
+    sigmas: Optional[List[float]] = None,
+    **kwargs,
+):
+    r"""
+    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
+    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
+
+    Args:
+        scheduler (`SchedulerMixin`):
+            The scheduler to get timesteps from.
+        num_inference_steps (`int`):
+            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
+            must be `None`.
+        device (`str` or `torch.device`, *optional*):
+            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
+        timesteps (`List[int]`, *optional*):
+            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
+            `num_inference_steps` and `sigmas` must be `None`.
+        sigmas (`List[float]`, *optional*):
+            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
+            `num_inference_steps` and `timesteps` must be `None`.
+
+    Returns:
+        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
+        second element is the number of inference steps.
+    """
+    if timesteps is not None and sigmas is not None:
+        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
+    if timesteps is not None:
+        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accepts_timesteps:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" timestep schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    elif sigmas is not None:
+        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accept_sigmas:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" sigmas schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    else:
+        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+    return timesteps, num_inference_steps
+
+
+def retrieve_latents(
+    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
+):
+    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
+        return encoder_output.latent_dist.sample(generator)
+    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
+        return encoder_output.latent_dist.mode()
+    elif hasattr(encoder_output, "latents"):
+        return encoder_output.latents
+    else:
+        raise AttributeError("Could not access latents of provided encoder_output")
+
+
+class SDMixin:
+    def _encode_prompt(
+        self,
+        prompt,
+        device,
+        num_images_per_prompt,
+        do_classifier_free_guidance,
+        negative_prompt=None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
+        lora_scale: Optional[float] = None,
+        **kwargs,
+    ):
+        deprecation_message = "`_encode_prompt()` is deprecated and it will be removed in a future version. Use `encode_prompt()` instead. Also, be aware that the output format changed from a concatenated tensor to a tuple."
+        deprecate("_encode_prompt()", "1.0.0", deprecation_message, standard_warn=False)
+
+        prompt_embeds_tuple = self.encode_prompt(
+            prompt=prompt,
+            device=device,
+            num_images_per_prompt=num_images_per_prompt,
+            do_classifier_free_guidance=do_classifier_free_guidance,
+            negative_prompt=negative_prompt,
+            prompt_embeds=prompt_embeds,
+            negative_prompt_embeds=negative_prompt_embeds,
+            lora_scale=lora_scale,
+            **kwargs,
+        )
+
+        # concatenate for backwards comp
+        prompt_embeds = torch.cat([prompt_embeds_tuple[1], prompt_embeds_tuple[0]])
+
+        return prompt_embeds
+
+    def encode_prompt(
+        self,
+        prompt,
+        device,
+        num_images_per_prompt,
+        do_classifier_free_guidance,
+        negative_prompt=None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
+        lora_scale: Optional[float] = None,
+        clip_skip: Optional[int] = None,
+    ):
+        r"""
+        Encodes the prompt into text encoder hidden states.
+
+        Args:
+            prompt (`str` or `List[str]`, *optional*):
+                prompt to be encoded
+            device: (`torch.device`):
+                torch device
+            num_images_per_prompt (`int`):
+                number of images that should be generated per prompt
+            do_classifier_free_guidance (`bool`):
+                whether to use classifier free guidance or not
+            negative_prompt (`str` or `List[str]`, *optional*):
+                The prompt or prompts not to guide the image generation. If not defined, one has to pass
+                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
+                less than `1`).
+            prompt_embeds (`torch.Tensor`, *optional*):
+                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
+                provided, text embeddings will be generated from `prompt` input argument.
+            negative_prompt_embeds (`torch.Tensor`, *optional*):
+                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
+                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
+                argument.
+            lora_scale (`float`, *optional*):
+                A LoRA scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
+            clip_skip (`int`, *optional*):
+                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
+                the output of the pre-final layer will be used for computing the prompt embeddings.
+        """
+        # set lora scale so that monkey patched LoRA
+        # function of text encoder can correctly access it
+        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
+            self._lora_scale = lora_scale
+
+            # dynamically adjust the LoRA scale
+            if not USE_PEFT_BACKEND:
+                adjust_lora_scale_text_encoder(self.text_encoder, lora_scale)
+            else:
+                scale_lora_layers(self.text_encoder, lora_scale)
+
+        if prompt is not None and isinstance(prompt, str):
+            batch_size = 1
+        elif prompt is not None and isinstance(prompt, list):
+            batch_size = len(prompt)
+        else:
+            batch_size = prompt_embeds.shape[0]
+
+        if prompt_embeds is None:
+            # textual inversion: process multi-vector tokens if necessary
+            if isinstance(self, TextualInversionLoaderMixin):
+                prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
+
+            text_inputs = self.tokenizer(
+                prompt,
+                padding="max_length",
+                max_length=self.tokenizer.model_max_length,
+                truncation=True,
+                return_tensors="pt",
+            )
+            text_input_ids = text_inputs.input_ids
+            untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
+
+            if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
+                text_input_ids, untruncated_ids
+            ):
+                removed_text = self.tokenizer.batch_decode(
+                    untruncated_ids[:, self.tokenizer.model_max_length - 1 : -1]
+                )
+                logger.warning(
+                    "The following part of your input was truncated because CLIP can only handle sequences up to"
+                    f" {self.tokenizer.model_max_length} tokens: {removed_text}"
+                )
+
+            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
+                attention_mask = text_inputs.attention_mask.to(device)
+            else:
+                attention_mask = None
+
+            if clip_skip is None:
+                prompt_embeds = self.text_encoder(text_input_ids.to(device), attention_mask=attention_mask)
+                prompt_embeds = prompt_embeds[0]
+            else:
+                prompt_embeds = self.text_encoder(
+                    text_input_ids.to(device), attention_mask=attention_mask, output_hidden_states=True
+                )
+                # Access the `hidden_states` first, that contains a tuple of
+                # all the hidden states from the encoder layers. Then index into
+                # the tuple to access the hidden states from the desired layer.
+                prompt_embeds = prompt_embeds[-1][-(clip_skip + 1)]
+                # We also need to apply the final LayerNorm here to not mess with the
+                # representations. The `last_hidden_states` that we typically use for
+                # obtaining the final prompt representations passes through the LayerNorm
+                # layer.
+                prompt_embeds = self.text_encoder.text_model.final_layer_norm(prompt_embeds)
+
+        if self.text_encoder is not None:
+            prompt_embeds_dtype = self.text_encoder.dtype
+        elif self.unet is not None:
+            prompt_embeds_dtype = self.unet.dtype
+        else:
+            prompt_embeds_dtype = prompt_embeds.dtype
+
+        prompt_embeds = prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
+
+        bs_embed, seq_len, _ = prompt_embeds.shape
+        # duplicate text embeddings for each generation per prompt, using mps friendly method
+        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
+        prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)
+
+        # get unconditional embeddings for classifier free guidance
+        if do_classifier_free_guidance and negative_prompt_embeds is None:
+            uncond_tokens: List[str]
+            if negative_prompt is None:
+                uncond_tokens = [""] * batch_size
+            elif prompt is not None and type(prompt) is not type(negative_prompt):
+                raise TypeError(
+                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
+                    f" {type(prompt)}."
+                )
+            elif isinstance(negative_prompt, str):
+                uncond_tokens = [negative_prompt]
+            elif batch_size != len(negative_prompt):
+                raise ValueError(
+                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
+                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
+                    " the batch size of `prompt`."
+                )
+            else:
+                uncond_tokens = negative_prompt
+
+            # textual inversion: process multi-vector tokens if necessary
+            if isinstance(self, TextualInversionLoaderMixin):
+                uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)
+
+            max_length = prompt_embeds.shape[1]
+            uncond_input = self.tokenizer(
+                uncond_tokens,
+                padding="max_length",
+                max_length=max_length,
+                truncation=True,
+                return_tensors="pt",
+            )
+
+            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
+                attention_mask = uncond_input.attention_mask.to(device)
+            else:
+                attention_mask = None
+
+            negative_prompt_embeds = self.text_encoder(
+                uncond_input.input_ids.to(device),
+                attention_mask=attention_mask,
+            )
+            negative_prompt_embeds = negative_prompt_embeds[0]
+
+        if do_classifier_free_guidance:
+            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
+            seq_len = negative_prompt_embeds.shape[1]
+
+            negative_prompt_embeds = negative_prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
+
+            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
+            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
+
+        if self.text_encoder is not None:
+            if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
+                # Retrieve the original scale by scaling back the LoRA layers
+                unscale_lora_layers(self.text_encoder, lora_scale)
+
+        return prompt_embeds, negative_prompt_embeds
+
+    def run_safety_checker(self, image, device, dtype):
+        if self.safety_checker is None:
+            has_nsfw_concept = None
+        else:
+            if torch.is_tensor(image):
+                feature_extractor_input = self.image_processor.postprocess(image, output_type="pil")
+            else:
+                feature_extractor_input = self.image_processor.numpy_to_pil(image)
+            safety_checker_input = self.feature_extractor(feature_extractor_input, return_tensors="pt").to(device)
+            image, has_nsfw_concept = self.safety_checker(
+                images=image, clip_input=safety_checker_input.pixel_values.to(dtype)
+            )
+        return image, has_nsfw_concept
+
+    def decode_latents(self, latents):
+        deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
+        deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
+
+        latents = 1 / self.vae.config.scaling_factor * latents
+        image = self.vae.decode(latents, return_dict=False)[0]
+        image = (image / 2 + 0.5).clamp(0, 1)
+        # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
+        image = image.cpu().permute(0, 2, 3, 1).float().numpy()
+        return image
+
+    def prepare_extra_step_kwargs(self, generator, eta):
+        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
+        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
+        # eta corresponds to η in DDIM paper: https://huggingface.co/papers/2010.02502
+        # and should be between [0, 1]
+
+        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
+        extra_step_kwargs = {}
+        if accepts_eta:
+            extra_step_kwargs["eta"] = eta
+
+        # check if the scheduler accepts generator
+        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
+        if accepts_generator:
+            extra_step_kwargs["generator"] = generator
+        return extra_step_kwargs
+
+    @property
+    def guidance_scale(self):
+        return self._guidance_scale
+
+    @property
+    def guidance_rescale(self):
+        return self._guidance_rescale
+
+    @property
+    def clip_skip(self):
+        return self._clip_skip
+
+    # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
+    # of the Imagen paper: https://huggingface.co/papers/2205.11487 . `guidance_scale = 1`
+    # corresponds to doing no classifier free guidance.
+    @property
+    def do_classifier_free_guidance(self):
+        return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
+
+    @property
+    def cross_attention_kwargs(self):
+        return self._cross_attention_kwargs
+
+    @property
+    def num_timesteps(self):
+        return self._num_timesteps
+
+    @property
+    def interrupt(self):
+        return self._interrupt
diff --git a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_unclip.py b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_unclip.py
index a134244e3ee4..c9f88e9d573b 100644
--- a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_unclip.py
+++ b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_unclip.py
@@ -23,19 +23,11 @@
 from ...loaders import StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, PriorTransformer, UNet2DConditionModel
 from ...models.embeddings import get_timestep_embedding
-from ...models.lora import adjust_lora_scale_text_encoder
 from ...schedulers import KarrasDiffusionSchedulers
-from ...utils import (
-    USE_PEFT_BACKEND,
-    deprecate,
-    is_torch_xla_available,
-    logging,
-    replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
-)
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput, StableDiffusionMixin
+from .pipeline_stable_diffusion_utils import SDMixin
 from .stable_unclip_image_normalizer import StableUnCLIPImageNormalizer
 
 
@@ -68,7 +60,7 @@
 
 
 class StableUnCLIPPipeline(
-    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, StableDiffusionLoraLoaderMixin
+    DiffusionPipeline, StableDiffusionMixin, SDMixin, TextualInversionLoaderMixin, StableDiffusionLoraLoaderMixin
 ):
     """
     Pipeline for text-to-image generation using stable unCLIP.
@@ -260,235 +252,7 @@ def _encode_prior_prompt(
 
         return prompt_embeds, text_enc_hid_states, text_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
-    def _encode_prompt(
-        self,
-        prompt,
-        device,
-        num_images_per_prompt,
-        do_classifier_free_guidance,
-        negative_prompt=None,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        negative_prompt_embeds: Optional[torch.Tensor] = None,
-        lora_scale: Optional[float] = None,
-        **kwargs,
-    ):
-        deprecation_message = "`_encode_prompt()` is deprecated and it will be removed in a future version. Use `encode_prompt()` instead. Also, be aware that the output format changed from a concatenated tensor to a tuple."
-        deprecate("_encode_prompt()", "1.0.0", deprecation_message, standard_warn=False)
-
-        prompt_embeds_tuple = self.encode_prompt(
-            prompt=prompt,
-            device=device,
-            num_images_per_prompt=num_images_per_prompt,
-            do_classifier_free_guidance=do_classifier_free_guidance,
-            negative_prompt=negative_prompt,
-            prompt_embeds=prompt_embeds,
-            negative_prompt_embeds=negative_prompt_embeds,
-            lora_scale=lora_scale,
-            **kwargs,
-        )
-
-        # concatenate for backwards comp
-        prompt_embeds = torch.cat([prompt_embeds_tuple[1], prompt_embeds_tuple[0]])
-
-        return prompt_embeds
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt,
-        device,
-        num_images_per_prompt,
-        do_classifier_free_guidance,
-        negative_prompt=None,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        negative_prompt_embeds: Optional[torch.Tensor] = None,
-        lora_scale: Optional[float] = None,
-        clip_skip: Optional[int] = None,
-    ):
-        r"""
-        Encodes the prompt into text encoder hidden states.
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            do_classifier_free_guidance (`bool`):
-                whether to use classifier free guidance or not
-            negative_prompt (`str` or `List[str]`, *optional*):
-                The prompt or prompts not to guide the image generation. If not defined, one has to pass
-                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
-                less than `1`).
-            prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
-                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
-                argument.
-            lora_scale (`float`, *optional*):
-                A LoRA scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-            clip_skip (`int`, *optional*):
-                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
-                the output of the pre-final layer will be used for computing the prompt embeddings.
-        """
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if not USE_PEFT_BACKEND:
-                adjust_lora_scale_text_encoder(self.text_encoder, lora_scale)
-            else:
-                scale_lora_layers(self.text_encoder, lora_scale)
-
-        if prompt is not None and isinstance(prompt, str):
-            batch_size = 1
-        elif prompt is not None and isinstance(prompt, list):
-            batch_size = len(prompt)
-        else:
-            batch_size = prompt_embeds.shape[0]
-
-        if prompt_embeds is None:
-            # textual inversion: process multi-vector tokens if necessary
-            if isinstance(self, TextualInversionLoaderMixin):
-                prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-            text_inputs = self.tokenizer(
-                prompt,
-                padding="max_length",
-                max_length=self.tokenizer.model_max_length,
-                truncation=True,
-                return_tensors="pt",
-            )
-            text_input_ids = text_inputs.input_ids
-            untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-
-            if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
-                text_input_ids, untruncated_ids
-            ):
-                removed_text = self.tokenizer.batch_decode(
-                    untruncated_ids[:, self.tokenizer.model_max_length - 1 : -1]
-                )
-                logger.warning(
-                    "The following part of your input was truncated because CLIP can only handle sequences up to"
-                    f" {self.tokenizer.model_max_length} tokens: {removed_text}"
-                )
-
-            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
-                attention_mask = text_inputs.attention_mask.to(device)
-            else:
-                attention_mask = None
-
-            if clip_skip is None:
-                prompt_embeds = self.text_encoder(text_input_ids.to(device), attention_mask=attention_mask)
-                prompt_embeds = prompt_embeds[0]
-            else:
-                prompt_embeds = self.text_encoder(
-                    text_input_ids.to(device), attention_mask=attention_mask, output_hidden_states=True
-                )
-                # Access the `hidden_states` first, that contains a tuple of
-                # all the hidden states from the encoder layers. Then index into
-                # the tuple to access the hidden states from the desired layer.
-                prompt_embeds = prompt_embeds[-1][-(clip_skip + 1)]
-                # We also need to apply the final LayerNorm here to not mess with the
-                # representations. The `last_hidden_states` that we typically use for
-                # obtaining the final prompt representations passes through the LayerNorm
-                # layer.
-                prompt_embeds = self.text_encoder.text_model.final_layer_norm(prompt_embeds)
-
-        if self.text_encoder is not None:
-            prompt_embeds_dtype = self.text_encoder.dtype
-        elif self.unet is not None:
-            prompt_embeds_dtype = self.unet.dtype
-        else:
-            prompt_embeds_dtype = prompt_embeds.dtype
-
-        prompt_embeds = prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
-
-        bs_embed, seq_len, _ = prompt_embeds.shape
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)
-
-        # get unconditional embeddings for classifier free guidance
-        if do_classifier_free_guidance and negative_prompt_embeds is None:
-            uncond_tokens: List[str]
-            if negative_prompt is None:
-                uncond_tokens = [""] * batch_size
-            elif prompt is not None and type(prompt) is not type(negative_prompt):
-                raise TypeError(
-                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
-                    f" {type(prompt)}."
-                )
-            elif isinstance(negative_prompt, str):
-                uncond_tokens = [negative_prompt]
-            elif batch_size != len(negative_prompt):
-                raise ValueError(
-                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
-                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
-                    " the batch size of `prompt`."
-                )
-            else:
-                uncond_tokens = negative_prompt
-
-            # textual inversion: process multi-vector tokens if necessary
-            if isinstance(self, TextualInversionLoaderMixin):
-                uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)
-
-            max_length = prompt_embeds.shape[1]
-            uncond_input = self.tokenizer(
-                uncond_tokens,
-                padding="max_length",
-                max_length=max_length,
-                truncation=True,
-                return_tensors="pt",
-            )
-
-            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
-                attention_mask = uncond_input.attention_mask.to(device)
-            else:
-                attention_mask = None
-
-            negative_prompt_embeds = self.text_encoder(
-                uncond_input.input_ids.to(device),
-                attention_mask=attention_mask,
-            )
-            negative_prompt_embeds = negative_prompt_embeds[0]
-
-        if do_classifier_free_guidance:
-            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
-            seq_len = negative_prompt_embeds.shape[1]
-
-            negative_prompt_embeds = negative_prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
-
-            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
-            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        if self.text_encoder is not None:
-            if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        return prompt_embeds, negative_prompt_embeds
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
-    def decode_latents(self, latents):
-        deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
-        deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
-
-        latents = 1 / self.vae.config.scaling_factor * latents
-        image = self.vae.decode(latents, return_dict=False)[0]
-        image = (image / 2 + 0.5).clamp(0, 1)
-        # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
-        image = image.cpu().permute(0, 2, 3, 1).float().numpy()
-        return image
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs with prepare_extra_step_kwargs->prepare_prior_extra_step_kwargs, scheduler->prior_scheduler
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs with prepare_extra_step_kwargs->prepare_prior_extra_step_kwargs, scheduler->prior_scheduler
     def prepare_prior_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the prior_scheduler step, since not all prior_schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other prior_schedulers.
@@ -506,24 +270,6 @@ def prepare_prior_extra_step_kwargs(self, generator, eta):
             extra_step_kwargs["generator"] = generator
         return extra_step_kwargs
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
-    def prepare_extra_step_kwargs(self, generator, eta):
-        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
-        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
-        # eta corresponds to η in DDIM paper: https://huggingface.co/papers/2010.02502
-        # and should be between [0, 1]
-
-        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        extra_step_kwargs = {}
-        if accepts_eta:
-            extra_step_kwargs["eta"] = eta
-
-        # check if the scheduler accepts generator
-        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        if accepts_generator:
-            extra_step_kwargs["generator"] = generator
-        return extra_step_kwargs
-
     def check_inputs(
         self,
         prompt,
diff --git a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_unclip_img2img.py b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_unclip_img2img.py
index abb4cc3a05d5..f6b6a2b74fd8 100644
--- a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_unclip_img2img.py
+++ b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_unclip_img2img.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
 import PIL.Image
@@ -23,19 +22,11 @@
 from ...loaders import StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, UNet2DConditionModel
 from ...models.embeddings import get_timestep_embedding
-from ...models.lora import adjust_lora_scale_text_encoder
 from ...schedulers import KarrasDiffusionSchedulers
-from ...utils import (
-    USE_PEFT_BACKEND,
-    deprecate,
-    is_torch_xla_available,
-    logging,
-    replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
-)
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput, StableDiffusionMixin
+from .pipeline_stable_diffusion_utils import SDMixin
 from .stable_unclip_image_normalizer import StableUnCLIPImageNormalizer
 
 
@@ -79,7 +70,7 @@
 
 
 class StableUnCLIPImg2ImgPipeline(
-    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, StableDiffusionLoraLoaderMixin
+    DiffusionPipeline, StableDiffusionMixin, SDMixin, TextualInversionLoaderMixin, StableDiffusionLoraLoaderMixin
 ):
     """
     Pipeline for text-guided image-to-image generation using stable unCLIP.
@@ -167,39 +158,6 @@ def __init__(
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
-    def _encode_prompt(
-        self,
-        prompt,
-        device,
-        num_images_per_prompt,
-        do_classifier_free_guidance,
-        negative_prompt=None,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        negative_prompt_embeds: Optional[torch.Tensor] = None,
-        lora_scale: Optional[float] = None,
-        **kwargs,
-    ):
-        deprecation_message = "`_encode_prompt()` is deprecated and it will be removed in a future version. Use `encode_prompt()` instead. Also, be aware that the output format changed from a concatenated tensor to a tuple."
-        deprecate("_encode_prompt()", "1.0.0", deprecation_message, standard_warn=False)
-
-        prompt_embeds_tuple = self.encode_prompt(
-            prompt=prompt,
-            device=device,
-            num_images_per_prompt=num_images_per_prompt,
-            do_classifier_free_guidance=do_classifier_free_guidance,
-            negative_prompt=negative_prompt,
-            prompt_embeds=prompt_embeds,
-            negative_prompt_embeds=negative_prompt_embeds,
-            lora_scale=lora_scale,
-            **kwargs,
-        )
-
-        # concatenate for backwards comp
-        prompt_embeds = torch.cat([prompt_embeds_tuple[1], prompt_embeds_tuple[0]])
-
-        return prompt_embeds
-
     def _encode_image(
         self,
         image,
@@ -255,219 +213,6 @@ def _encode_image(
 
         return image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt,
-        device,
-        num_images_per_prompt,
-        do_classifier_free_guidance,
-        negative_prompt=None,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        negative_prompt_embeds: Optional[torch.Tensor] = None,
-        lora_scale: Optional[float] = None,
-        clip_skip: Optional[int] = None,
-    ):
-        r"""
-        Encodes the prompt into text encoder hidden states.
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            do_classifier_free_guidance (`bool`):
-                whether to use classifier free guidance or not
-            negative_prompt (`str` or `List[str]`, *optional*):
-                The prompt or prompts not to guide the image generation. If not defined, one has to pass
-                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
-                less than `1`).
-            prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
-                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
-                argument.
-            lora_scale (`float`, *optional*):
-                A LoRA scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-            clip_skip (`int`, *optional*):
-                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
-                the output of the pre-final layer will be used for computing the prompt embeddings.
-        """
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if not USE_PEFT_BACKEND:
-                adjust_lora_scale_text_encoder(self.text_encoder, lora_scale)
-            else:
-                scale_lora_layers(self.text_encoder, lora_scale)
-
-        if prompt is not None and isinstance(prompt, str):
-            batch_size = 1
-        elif prompt is not None and isinstance(prompt, list):
-            batch_size = len(prompt)
-        else:
-            batch_size = prompt_embeds.shape[0]
-
-        if prompt_embeds is None:
-            # textual inversion: process multi-vector tokens if necessary
-            if isinstance(self, TextualInversionLoaderMixin):
-                prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
-
-            text_inputs = self.tokenizer(
-                prompt,
-                padding="max_length",
-                max_length=self.tokenizer.model_max_length,
-                truncation=True,
-                return_tensors="pt",
-            )
-            text_input_ids = text_inputs.input_ids
-            untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-
-            if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
-                text_input_ids, untruncated_ids
-            ):
-                removed_text = self.tokenizer.batch_decode(
-                    untruncated_ids[:, self.tokenizer.model_max_length - 1 : -1]
-                )
-                logger.warning(
-                    "The following part of your input was truncated because CLIP can only handle sequences up to"
-                    f" {self.tokenizer.model_max_length} tokens: {removed_text}"
-                )
-
-            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
-                attention_mask = text_inputs.attention_mask.to(device)
-            else:
-                attention_mask = None
-
-            if clip_skip is None:
-                prompt_embeds = self.text_encoder(text_input_ids.to(device), attention_mask=attention_mask)
-                prompt_embeds = prompt_embeds[0]
-            else:
-                prompt_embeds = self.text_encoder(
-                    text_input_ids.to(device), attention_mask=attention_mask, output_hidden_states=True
-                )
-                # Access the `hidden_states` first, that contains a tuple of
-                # all the hidden states from the encoder layers. Then index into
-                # the tuple to access the hidden states from the desired layer.
-                prompt_embeds = prompt_embeds[-1][-(clip_skip + 1)]
-                # We also need to apply the final LayerNorm here to not mess with the
-                # representations. The `last_hidden_states` that we typically use for
-                # obtaining the final prompt representations passes through the LayerNorm
-                # layer.
-                prompt_embeds = self.text_encoder.text_model.final_layer_norm(prompt_embeds)
-
-        if self.text_encoder is not None:
-            prompt_embeds_dtype = self.text_encoder.dtype
-        elif self.unet is not None:
-            prompt_embeds_dtype = self.unet.dtype
-        else:
-            prompt_embeds_dtype = prompt_embeds.dtype
-
-        prompt_embeds = prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
-
-        bs_embed, seq_len, _ = prompt_embeds.shape
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)
-
-        # get unconditional embeddings for classifier free guidance
-        if do_classifier_free_guidance and negative_prompt_embeds is None:
-            uncond_tokens: List[str]
-            if negative_prompt is None:
-                uncond_tokens = [""] * batch_size
-            elif prompt is not None and type(prompt) is not type(negative_prompt):
-                raise TypeError(
-                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
-                    f" {type(prompt)}."
-                )
-            elif isinstance(negative_prompt, str):
-                uncond_tokens = [negative_prompt]
-            elif batch_size != len(negative_prompt):
-                raise ValueError(
-                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
-                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
-                    " the batch size of `prompt`."
-                )
-            else:
-                uncond_tokens = negative_prompt
-
-            # textual inversion: process multi-vector tokens if necessary
-            if isinstance(self, TextualInversionLoaderMixin):
-                uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)
-
-            max_length = prompt_embeds.shape[1]
-            uncond_input = self.tokenizer(
-                uncond_tokens,
-                padding="max_length",
-                max_length=max_length,
-                truncation=True,
-                return_tensors="pt",
-            )
-
-            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
-                attention_mask = uncond_input.attention_mask.to(device)
-            else:
-                attention_mask = None
-
-            negative_prompt_embeds = self.text_encoder(
-                uncond_input.input_ids.to(device),
-                attention_mask=attention_mask,
-            )
-            negative_prompt_embeds = negative_prompt_embeds[0]
-
-        if do_classifier_free_guidance:
-            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
-            seq_len = negative_prompt_embeds.shape[1]
-
-            negative_prompt_embeds = negative_prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
-
-            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
-            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        if self.text_encoder is not None:
-            if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        return prompt_embeds, negative_prompt_embeds
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
-    def decode_latents(self, latents):
-        deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
-        deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
-
-        latents = 1 / self.vae.config.scaling_factor * latents
-        image = self.vae.decode(latents, return_dict=False)[0]
-        image = (image / 2 + 0.5).clamp(0, 1)
-        # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
-        image = image.cpu().permute(0, 2, 3, 1).float().numpy()
-        return image
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
-    def prepare_extra_step_kwargs(self, generator, eta):
-        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
-        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
-        # eta corresponds to η in DDIM paper: https://huggingface.co/papers/2010.02502
-        # and should be between [0, 1]
-
-        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        extra_step_kwargs = {}
-        if accepts_eta:
-            extra_step_kwargs["eta"] = eta
-
-        # check if the scheduler accepts generator
-        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        if accepts_generator:
-            extra_step_kwargs["generator"] = generator
-        return extra_step_kwargs
-
     def check_inputs(
         self,
         prompt,
diff --git a/src/diffusers/pipelines/stable_diffusion_3/pipeline_stable_diffusion_3.py b/src/diffusers/pipelines/stable_diffusion_3/pipeline_stable_diffusion_3.py
index 660d9801df56..5d49fdd24635 100644
--- a/src/diffusers/pipelines/stable_diffusion_3/pipeline_stable_diffusion_3.py
+++ b/src/diffusers/pipelines/stable_diffusion_3/pipeline_stable_diffusion_3.py
@@ -71,7 +71,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -85,7 +85,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/stable_diffusion_3/pipeline_stable_diffusion_3_img2img.py b/src/diffusers/pipelines/stable_diffusion_3/pipeline_stable_diffusion_3_img2img.py
index 9b11bc8781e7..e9c0667a2c84 100644
--- a/src/diffusers/pipelines/stable_diffusion_3/pipeline_stable_diffusion_3_img2img.py
+++ b/src/diffusers/pipelines/stable_diffusion_3/pipeline_stable_diffusion_3_img2img.py
@@ -77,7 +77,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -91,7 +91,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -105,7 +105,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/stable_diffusion_3/pipeline_stable_diffusion_3_inpaint.py b/src/diffusers/pipelines/stable_diffusion_3/pipeline_stable_diffusion_3_inpaint.py
index b947cbff0914..995d08c4d7ff 100644
--- a/src/diffusers/pipelines/stable_diffusion_3/pipeline_stable_diffusion_3_inpaint.py
+++ b/src/diffusers/pipelines/stable_diffusion_3/pipeline_stable_diffusion_3_inpaint.py
@@ -76,7 +76,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -90,7 +90,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -104,7 +104,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/stable_diffusion_attend_and_excite/pipeline_stable_diffusion_attend_and_excite.py b/src/diffusers/pipelines/stable_diffusion_attend_and_excite/pipeline_stable_diffusion_attend_and_excite.py
index a1ff99b6aa34..aee3ff1f2a92 100644
--- a/src/diffusers/pipelines/stable_diffusion_attend_and_excite/pipeline_stable_diffusion_attend_and_excite.py
+++ b/src/diffusers/pipelines/stable_diffusion_attend_and_excite/pipeline_stable_diffusion_attend_and_excite.py
@@ -259,7 +259,7 @@ def __init__(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -292,7 +292,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -475,7 +475,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -490,7 +490,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -502,7 +502,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/stable_diffusion_diffedit/pipeline_stable_diffusion_diffedit.py b/src/diffusers/pipelines/stable_diffusion_diffedit/pipeline_stable_diffusion_diffedit.py
index 65c25ffbe492..12e83d21ff84 100644
--- a/src/diffusers/pipelines/stable_diffusion_diffedit/pipeline_stable_diffusion_diffedit.py
+++ b/src/diffusers/pipelines/stable_diffusion_diffedit/pipeline_stable_diffusion_diffedit.py
@@ -385,7 +385,7 @@ def __init__(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -418,7 +418,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -601,7 +601,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -616,7 +616,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -634,7 +634,7 @@ def prepare_extra_step_kwargs(self, generator, eta):
             extra_step_kwargs["generator"] = generator
         return extra_step_kwargs
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
diff --git a/src/diffusers/pipelines/stable_diffusion_gligen/pipeline_stable_diffusion_gligen.py b/src/diffusers/pipelines/stable_diffusion_gligen/pipeline_stable_diffusion_gligen.py
index 78b026684cfa..86f8482f4085 100644
--- a/src/diffusers/pipelines/stable_diffusion_gligen/pipeline_stable_diffusion_gligen.py
+++ b/src/diffusers/pipelines/stable_diffusion_gligen/pipeline_stable_diffusion_gligen.py
@@ -183,7 +183,7 @@ def __init__(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -216,7 +216,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -399,7 +399,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
diff --git a/src/diffusers/pipelines/stable_diffusion_gligen/pipeline_stable_diffusion_gligen_text_image.py b/src/diffusers/pipelines/stable_diffusion_gligen/pipeline_stable_diffusion_gligen_text_image.py
index 05cbad139d92..b01a98ce4049 100644
--- a/src/diffusers/pipelines/stable_diffusion_gligen/pipeline_stable_diffusion_gligen_text_image.py
+++ b/src/diffusers/pipelines/stable_diffusion_gligen/pipeline_stable_diffusion_gligen_text_image.py
@@ -247,7 +247,7 @@ def __init__(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -430,7 +430,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -445,7 +445,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/stable_diffusion_k_diffusion/pipeline_stable_diffusion_k_diffusion.py b/src/diffusers/pipelines/stable_diffusion_k_diffusion/pipeline_stable_diffusion_k_diffusion.py
index feebd6adf8f8..27143bb3a4b0 100755
--- a/src/diffusers/pipelines/stable_diffusion_k_diffusion/pipeline_stable_diffusion_k_diffusion.py
+++ b/src/diffusers/pipelines/stable_diffusion_k_diffusion/pipeline_stable_diffusion_k_diffusion.py
@@ -165,7 +165,7 @@ def set_scheduler(self, scheduler_type: str):
 
             raise ValueError(f"Invalid scheduler type {scheduler_type}. Please choose one of {valid_samplers}.")
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -198,7 +198,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -381,7 +381,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -396,7 +396,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
diff --git a/src/diffusers/pipelines/stable_diffusion_k_diffusion/pipeline_stable_diffusion_xl_k_diffusion.py b/src/diffusers/pipelines/stable_diffusion_k_diffusion/pipeline_stable_diffusion_xl_k_diffusion.py
index 766ca37d8142..1e1787016361 100644
--- a/src/diffusers/pipelines/stable_diffusion_k_diffusion/pipeline_stable_diffusion_xl_k_diffusion.py
+++ b/src/diffusers/pipelines/stable_diffusion_k_diffusion/pipeline_stable_diffusion_xl_k_diffusion.py
@@ -33,11 +33,7 @@
     TextualInversionLoaderMixin,
 )
 from ...models import AutoencoderKL, UNet2DConditionModel
-from ...models.attention_processor import (
-    AttnProcessor2_0,
-    FusedAttnProcessor2_0,
-    XFormersAttnProcessor,
-)
+from ...models.attention_processor import AttnProcessor2_0, FusedAttnProcessor2_0, XFormersAttnProcessor
 from ...models.lora import adjust_lora_scale_text_encoder
 from ...schedulers import KarrasDiffusionSchedulers, LMSDiscreteScheduler
 from ...utils import (
@@ -202,7 +198,7 @@ def set_scheduler(self, scheduler_type: str):
 
             raise ValueError(f"Invalid scheduler type {scheduler_type}. Please choose one of {valid_samplers}.")
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -543,17 +539,13 @@ def _get_add_time_ids(
         add_time_ids = torch.tensor([add_time_ids], dtype=dtype)
         return add_time_ids
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.upcast_vae
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.upcast_vae
     def upcast_vae(self):
         dtype = self.vae.dtype
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-                FusedAttnProcessor2_0,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor, FusedAttnProcessor2_0),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/pipelines/stable_diffusion_ldm3d/pipeline_stable_diffusion_ldm3d.py b/src/diffusers/pipelines/stable_diffusion_ldm3d/pipeline_stable_diffusion_ldm3d.py
index c32121c88c9b..61e95e05d4e5 100644
--- a/src/diffusers/pipelines/stable_diffusion_ldm3d/pipeline_stable_diffusion_ldm3d.py
+++ b/src/diffusers/pipelines/stable_diffusion_ldm3d/pipeline_stable_diffusion_ldm3d.py
@@ -68,7 +68,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -95,7 +95,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -270,7 +270,7 @@ def __init__(
         self.image_processor = VaeImageProcessorLDM3D(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -303,7 +303,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -572,7 +572,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/stable_diffusion_panorama/pipeline_stable_diffusion_panorama.py b/src/diffusers/pipelines/stable_diffusion_panorama/pipeline_stable_diffusion_panorama.py
index 295095947a12..f2b812a3d64d 100644
--- a/src/diffusers/pipelines/stable_diffusion_panorama/pipeline_stable_diffusion_panorama.py
+++ b/src/diffusers/pipelines/stable_diffusion_panorama/pipeline_stable_diffusion_panorama.py
@@ -68,7 +68,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -95,7 +95,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -246,7 +246,7 @@ def __init__(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -279,7 +279,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -533,7 +533,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -548,7 +548,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -586,7 +586,7 @@ def decode_latents_with_padding(self, latents: torch.Tensor, padding: int = 8) -
         image = image[..., padding_pix:-padding_pix]
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/stable_diffusion_safe/pipeline_stable_diffusion_safe.py b/src/diffusers/pipelines/stable_diffusion_safe/pipeline_stable_diffusion_safe.py
index d334107b0703..cf34d135b453 100644
--- a/src/diffusers/pipelines/stable_diffusion_safe/pipeline_stable_diffusion_safe.py
+++ b/src/diffusers/pipelines/stable_diffusion_safe/pipeline_stable_diffusion_safe.py
@@ -344,7 +344,7 @@ def run_safety_checker(self, image, device, dtype, enable_safety_guidance):
             flagged_images = None
         return image, has_nsfw_concept, flagged_images
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -356,7 +356,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/stable_diffusion_sag/pipeline_stable_diffusion_sag.py b/src/diffusers/pipelines/stable_diffusion_sag/pipeline_stable_diffusion_sag.py
index 48add535a81d..93ef2690c52d 100644
--- a/src/diffusers/pipelines/stable_diffusion_sag/pipeline_stable_diffusion_sag.py
+++ b/src/diffusers/pipelines/stable_diffusion_sag/pipeline_stable_diffusion_sag.py
@@ -174,7 +174,7 @@ def __init__(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -207,7 +207,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -449,7 +449,7 @@ def prepare_ip_adapter_image_embeds(
             image_embeds = ip_adapter_image_embeds
         return image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -464,7 +464,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -476,7 +476,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py b/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py
index b97cf6f1f6f8..f75bcef39986 100644
--- a/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py
+++ b/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Tuple, Union
 
 import torch
@@ -32,27 +31,19 @@
     StableDiffusionXLLoraLoaderMixin,
     TextualInversionLoaderMixin,
 )
-from ...models import AutoencoderKL, ImageProjection, UNet2DConditionModel
-from ...models.attention_processor import (
-    AttnProcessor2_0,
-    FusedAttnProcessor2_0,
-    XFormersAttnProcessor,
-)
-from ...models.lora import adjust_lora_scale_text_encoder
+from ...models import AutoencoderKL, UNet2DConditionModel
 from ...schedulers import KarrasDiffusionSchedulers
 from ...utils import (
-    USE_PEFT_BACKEND,
     deprecate,
     is_invisible_watermark_available,
     is_torch_xla_available,
     logging,
     replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
 )
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from .pipeline_output import StableDiffusionXLPipelineOutput
+from .pipeline_stable_diffusion_xl_utils import StableDiffusionXLMixin, rescale_noise_cfg, retrieve_timesteps
 
 
 if is_invisible_watermark_available():
@@ -85,95 +76,9 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
-def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
-    r"""
-    Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
-    Section 3.4 from [Common Diffusion Noise Schedules and Sample Steps are
-    Flawed](https://huggingface.co/papers/2305.08891).
-
-    Args:
-        noise_cfg (`torch.Tensor`):
-            The predicted noise tensor for the guided diffusion process.
-        noise_pred_text (`torch.Tensor`):
-            The predicted noise tensor for the text-guided diffusion process.
-        guidance_rescale (`float`, *optional*, defaults to 0.0):
-            A rescale factor applied to the noise predictions.
-
-    Returns:
-        noise_cfg (`torch.Tensor`): The rescaled noise prediction tensor.
-    """
-    std_text = noise_pred_text.std(dim=list(range(1, noise_pred_text.ndim)), keepdim=True)
-    std_cfg = noise_cfg.std(dim=list(range(1, noise_cfg.ndim)), keepdim=True)
-    # rescale the results from guidance (fixes overexposure)
-    noise_pred_rescaled = noise_cfg * (std_text / std_cfg)
-    # mix with the original results from guidance by factor guidance_rescale to avoid "plain looking" images
-    noise_cfg = guidance_rescale * noise_pred_rescaled + (1 - guidance_rescale) * noise_cfg
-    return noise_cfg
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
 class StableDiffusionXLPipeline(
     DiffusionPipeline,
+    StableDiffusionXLMixin,
     StableDiffusionMixin,
     FromSingleFileMixin,
     StableDiffusionXLLoraLoaderMixin,
@@ -285,333 +190,6 @@ def __init__(
         else:
             self.watermark = None
 
-    def encode_prompt(
-        self,
-        prompt: str,
-        prompt_2: Optional[str] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        do_classifier_free_guidance: bool = True,
-        negative_prompt: Optional[str] = None,
-        negative_prompt_2: Optional[str] = None,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        negative_prompt_embeds: Optional[torch.Tensor] = None,
-        pooled_prompt_embeds: Optional[torch.Tensor] = None,
-        negative_pooled_prompt_embeds: Optional[torch.Tensor] = None,
-        lora_scale: Optional[float] = None,
-        clip_skip: Optional[int] = None,
-    ):
-        r"""
-        Encodes the prompt into text encoder hidden states.
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                used in both text-encoders
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            do_classifier_free_guidance (`bool`):
-                whether to use classifier free guidance or not
-            negative_prompt (`str` or `List[str]`, *optional*):
-                The prompt or prompts not to guide the image generation. If not defined, one has to pass
-                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
-                less than `1`).
-            negative_prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts not to guide the image generation to be sent to `tokenizer_2` and
-                `text_encoder_2`. If not defined, `negative_prompt` is used in both text-encoders
-            prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
-                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
-                argument.
-            pooled_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            negative_pooled_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated negative pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
-                weighting. If not provided, pooled negative_prompt_embeds will be generated from `negative_prompt`
-                input argument.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-            clip_skip (`int`, *optional*):
-                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
-                the output of the pre-final layer will be used for computing the prompt embeddings.
-        """
-        device = device or self._execution_device
-
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionXLLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None:
-                if not USE_PEFT_BACKEND:
-                    adjust_lora_scale_text_encoder(self.text_encoder, lora_scale)
-                else:
-                    scale_lora_layers(self.text_encoder, lora_scale)
-
-            if self.text_encoder_2 is not None:
-                if not USE_PEFT_BACKEND:
-                    adjust_lora_scale_text_encoder(self.text_encoder_2, lora_scale)
-                else:
-                    scale_lora_layers(self.text_encoder_2, lora_scale)
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        if prompt is not None:
-            batch_size = len(prompt)
-        else:
-            batch_size = prompt_embeds.shape[0]
-
-        # Define tokenizers and text encoders
-        tokenizers = [self.tokenizer, self.tokenizer_2] if self.tokenizer is not None else [self.tokenizer_2]
-        text_encoders = (
-            [self.text_encoder, self.text_encoder_2] if self.text_encoder is not None else [self.text_encoder_2]
-        )
-
-        if prompt_embeds is None:
-            prompt_2 = prompt_2 or prompt
-            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
-
-            # textual inversion: process multi-vector tokens if necessary
-            prompt_embeds_list = []
-            prompts = [prompt, prompt_2]
-            for prompt, tokenizer, text_encoder in zip(prompts, tokenizers, text_encoders):
-                if isinstance(self, TextualInversionLoaderMixin):
-                    prompt = self.maybe_convert_prompt(prompt, tokenizer)
-
-                text_inputs = tokenizer(
-                    prompt,
-                    padding="max_length",
-                    max_length=tokenizer.model_max_length,
-                    truncation=True,
-                    return_tensors="pt",
-                )
-
-                text_input_ids = text_inputs.input_ids
-                untruncated_ids = tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-
-                if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
-                    text_input_ids, untruncated_ids
-                ):
-                    removed_text = tokenizer.batch_decode(untruncated_ids[:, tokenizer.model_max_length - 1 : -1])
-                    logger.warning(
-                        "The following part of your input was truncated because CLIP can only handle sequences up to"
-                        f" {tokenizer.model_max_length} tokens: {removed_text}"
-                    )
-
-                prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)
-
-                # We are only ALWAYS interested in the pooled output of the final text encoder
-                if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
-                    pooled_prompt_embeds = prompt_embeds[0]
-
-                if clip_skip is None:
-                    prompt_embeds = prompt_embeds.hidden_states[-2]
-                else:
-                    # "2" because SDXL always indexes from the penultimate layer.
-                    prompt_embeds = prompt_embeds.hidden_states[-(clip_skip + 2)]
-
-                prompt_embeds_list.append(prompt_embeds)
-
-            prompt_embeds = torch.concat(prompt_embeds_list, dim=-1)
-
-        # get unconditional embeddings for classifier free guidance
-        zero_out_negative_prompt = negative_prompt is None and self.config.force_zeros_for_empty_prompt
-        if do_classifier_free_guidance and negative_prompt_embeds is None and zero_out_negative_prompt:
-            negative_prompt_embeds = torch.zeros_like(prompt_embeds)
-            negative_pooled_prompt_embeds = torch.zeros_like(pooled_prompt_embeds)
-        elif do_classifier_free_guidance and negative_prompt_embeds is None:
-            negative_prompt = negative_prompt or ""
-            negative_prompt_2 = negative_prompt_2 or negative_prompt
-
-            # normalize str to list
-            negative_prompt = batch_size * [negative_prompt] if isinstance(negative_prompt, str) else negative_prompt
-            negative_prompt_2 = (
-                batch_size * [negative_prompt_2] if isinstance(negative_prompt_2, str) else negative_prompt_2
-            )
-
-            uncond_tokens: List[str]
-            if prompt is not None and type(prompt) is not type(negative_prompt):
-                raise TypeError(
-                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
-                    f" {type(prompt)}."
-                )
-            elif batch_size != len(negative_prompt):
-                raise ValueError(
-                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
-                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
-                    " the batch size of `prompt`."
-                )
-            else:
-                uncond_tokens = [negative_prompt, negative_prompt_2]
-
-            negative_prompt_embeds_list = []
-            for negative_prompt, tokenizer, text_encoder in zip(uncond_tokens, tokenizers, text_encoders):
-                if isinstance(self, TextualInversionLoaderMixin):
-                    negative_prompt = self.maybe_convert_prompt(negative_prompt, tokenizer)
-
-                max_length = prompt_embeds.shape[1]
-                uncond_input = tokenizer(
-                    negative_prompt,
-                    padding="max_length",
-                    max_length=max_length,
-                    truncation=True,
-                    return_tensors="pt",
-                )
-
-                negative_prompt_embeds = text_encoder(
-                    uncond_input.input_ids.to(device),
-                    output_hidden_states=True,
-                )
-
-                # We are only ALWAYS interested in the pooled output of the final text encoder
-                if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
-                    negative_pooled_prompt_embeds = negative_prompt_embeds[0]
-                negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]
-
-                negative_prompt_embeds_list.append(negative_prompt_embeds)
-
-            negative_prompt_embeds = torch.concat(negative_prompt_embeds_list, dim=-1)
-
-        if self.text_encoder_2 is not None:
-            prompt_embeds = prompt_embeds.to(dtype=self.text_encoder_2.dtype, device=device)
-        else:
-            prompt_embeds = prompt_embeds.to(dtype=self.unet.dtype, device=device)
-
-        bs_embed, seq_len, _ = prompt_embeds.shape
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)
-
-        if do_classifier_free_guidance:
-            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
-            seq_len = negative_prompt_embeds.shape[1]
-
-            if self.text_encoder_2 is not None:
-                negative_prompt_embeds = negative_prompt_embeds.to(dtype=self.text_encoder_2.dtype, device=device)
-            else:
-                negative_prompt_embeds = negative_prompt_embeds.to(dtype=self.unet.dtype, device=device)
-
-            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
-            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        pooled_prompt_embeds = pooled_prompt_embeds.repeat(1, num_images_per_prompt).view(
-            bs_embed * num_images_per_prompt, -1
-        )
-        if do_classifier_free_guidance:
-            negative_pooled_prompt_embeds = negative_pooled_prompt_embeds.repeat(1, num_images_per_prompt).view(
-                bs_embed * num_images_per_prompt, -1
-            )
-
-        if self.text_encoder is not None:
-            if isinstance(self, StableDiffusionXLLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        if self.text_encoder_2 is not None:
-            if isinstance(self, StableDiffusionXLLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder_2, lora_scale)
-
-        return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_image
-    def encode_image(self, image, device, num_images_per_prompt, output_hidden_states=None):
-        dtype = next(self.image_encoder.parameters()).dtype
-
-        if not isinstance(image, torch.Tensor):
-            image = self.feature_extractor(image, return_tensors="pt").pixel_values
-
-        image = image.to(device=device, dtype=dtype)
-        if output_hidden_states:
-            image_enc_hidden_states = self.image_encoder(image, output_hidden_states=True).hidden_states[-2]
-            image_enc_hidden_states = image_enc_hidden_states.repeat_interleave(num_images_per_prompt, dim=0)
-            uncond_image_enc_hidden_states = self.image_encoder(
-                torch.zeros_like(image), output_hidden_states=True
-            ).hidden_states[-2]
-            uncond_image_enc_hidden_states = uncond_image_enc_hidden_states.repeat_interleave(
-                num_images_per_prompt, dim=0
-            )
-            return image_enc_hidden_states, uncond_image_enc_hidden_states
-        else:
-            image_embeds = self.image_encoder(image).image_embeds
-            image_embeds = image_embeds.repeat_interleave(num_images_per_prompt, dim=0)
-            uncond_image_embeds = torch.zeros_like(image_embeds)
-
-            return image_embeds, uncond_image_embeds
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_ip_adapter_image_embeds
-    def prepare_ip_adapter_image_embeds(
-        self, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt, do_classifier_free_guidance
-    ):
-        image_embeds = []
-        if do_classifier_free_guidance:
-            negative_image_embeds = []
-        if ip_adapter_image_embeds is None:
-            if not isinstance(ip_adapter_image, list):
-                ip_adapter_image = [ip_adapter_image]
-
-            if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
-                raise ValueError(
-                    f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
-                )
-
-            for single_ip_adapter_image, image_proj_layer in zip(
-                ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
-            ):
-                output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
-                single_image_embeds, single_negative_image_embeds = self.encode_image(
-                    single_ip_adapter_image, device, 1, output_hidden_state
-                )
-
-                image_embeds.append(single_image_embeds[None, :])
-                if do_classifier_free_guidance:
-                    negative_image_embeds.append(single_negative_image_embeds[None, :])
-        else:
-            for single_image_embeds in ip_adapter_image_embeds:
-                if do_classifier_free_guidance:
-                    single_negative_image_embeds, single_image_embeds = single_image_embeds.chunk(2)
-                    negative_image_embeds.append(single_negative_image_embeds)
-                image_embeds.append(single_image_embeds)
-
-        ip_adapter_image_embeds = []
-        for i, single_image_embeds in enumerate(image_embeds):
-            single_image_embeds = torch.cat([single_image_embeds] * num_images_per_prompt, dim=0)
-            if do_classifier_free_guidance:
-                single_negative_image_embeds = torch.cat([negative_image_embeds[i]] * num_images_per_prompt, dim=0)
-                single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds], dim=0)
-
-            single_image_embeds = single_image_embeds.to(device=device)
-            ip_adapter_image_embeds.append(single_image_embeds)
-
-        return ip_adapter_image_embeds
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
-    def prepare_extra_step_kwargs(self, generator, eta):
-        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
-        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
-        # eta corresponds to η in DDIM paper: https://huggingface.co/papers/2010.02502
-        # and should be between [0, 1]
-
-        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        extra_step_kwargs = {}
-        if accepts_eta:
-            extra_step_kwargs["eta"] = eta
-
-        # check if the scheduler accepts generator
-        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        if accepts_generator:
-            extra_step_kwargs["generator"] = generator
-        return extra_step_kwargs
-
     def check_inputs(
         self,
         prompt,
@@ -749,90 +327,6 @@ def _get_add_time_ids(
         add_time_ids = torch.tensor([add_time_ids], dtype=dtype)
         return add_time_ids
 
-    def upcast_vae(self):
-        dtype = self.vae.dtype
-        self.vae.to(dtype=torch.float32)
-        use_torch_2_0_or_xformers = isinstance(
-            self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-                FusedAttnProcessor2_0,
-            ),
-        )
-        # if xformers or torch_2_0 is used attention block does not need
-        # to be in float32 which can save lots of memory
-        if use_torch_2_0_or_xformers:
-            self.vae.post_quant_conv.to(dtype)
-            self.vae.decoder.conv_in.to(dtype)
-            self.vae.decoder.mid_block.to(dtype)
-
-    # Copied from diffusers.pipelines.latent_consistency_models.pipeline_latent_consistency_text2img.LatentConsistencyModelPipeline.get_guidance_scale_embedding
-    def get_guidance_scale_embedding(
-        self, w: torch.Tensor, embedding_dim: int = 512, dtype: torch.dtype = torch.float32
-    ) -> torch.Tensor:
-        """
-        See https://github.com/google-research/vdm/blob/dc27b98a554f65cdc654b800da5aa1846545d41b/model_vdm.py#L298
-
-        Args:
-            w (`torch.Tensor`):
-                Generate embedding vectors with a specified guidance scale to subsequently enrich timestep embeddings.
-            embedding_dim (`int`, *optional*, defaults to 512):
-                Dimension of the embeddings to generate.
-            dtype (`torch.dtype`, *optional*, defaults to `torch.float32`):
-                Data type of the generated embeddings.
-
-        Returns:
-            `torch.Tensor`: Embedding vectors with shape `(len(w), embedding_dim)`.
-        """
-        assert len(w.shape) == 1
-        w = w * 1000.0
-
-        half_dim = embedding_dim // 2
-        emb = torch.log(torch.tensor(10000.0)) / (half_dim - 1)
-        emb = torch.exp(torch.arange(half_dim, dtype=dtype) * -emb)
-        emb = w.to(dtype)[:, None] * emb[None, :]
-        emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=1)
-        if embedding_dim % 2 == 1:  # zero pad
-            emb = torch.nn.functional.pad(emb, (0, 1))
-        assert emb.shape == (w.shape[0], embedding_dim)
-        return emb
-
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def guidance_rescale(self):
-        return self._guidance_rescale
-
-    @property
-    def clip_skip(self):
-        return self._clip_skip
-
-    # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
-    # of the Imagen paper: https://huggingface.co/papers/2205.11487 . `guidance_scale = 1`
-    # corresponds to doing no classifier free guidance.
-    @property
-    def do_classifier_free_guidance(self):
-        return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
-
-    @property
-    def cross_attention_kwargs(self):
-        return self._cross_attention_kwargs
-
-    @property
-    def denoising_end(self):
-        return self._denoising_end
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_img2img.py b/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_img2img.py
index 44e8f4fe4b54..d8b90791d606 100644
--- a/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_img2img.py
+++ b/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_img2img.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Tuple, Union
 
 import PIL.Image
@@ -33,26 +32,24 @@
     StableDiffusionXLLoraLoaderMixin,
     TextualInversionLoaderMixin,
 )
-from ...models import AutoencoderKL, ImageProjection, UNet2DConditionModel
-from ...models.attention_processor import (
-    AttnProcessor2_0,
-    XFormersAttnProcessor,
-)
-from ...models.lora import adjust_lora_scale_text_encoder
+from ...models import AutoencoderKL, UNet2DConditionModel
 from ...schedulers import KarrasDiffusionSchedulers
 from ...utils import (
-    USE_PEFT_BACKEND,
     deprecate,
     is_invisible_watermark_available,
     is_torch_xla_available,
     logging,
     replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
 )
 from ...utils.torch_utils import empty_device_cache, randn_tensor
 from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from .pipeline_output import StableDiffusionXLPipelineOutput
+from .pipeline_stable_diffusion_xl_utils import (
+    StableDiffusionXLMixin,
+    rescale_noise_cfg,
+    retrieve_latents,
+    retrieve_timesteps,
+)
 
 
 if is_invisible_watermark_available():
@@ -88,109 +85,9 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
-def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
-    r"""
-    Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
-    Section 3.4 from [Common Diffusion Noise Schedules and Sample Steps are
-    Flawed](https://huggingface.co/papers/2305.08891).
-
-    Args:
-        noise_cfg (`torch.Tensor`):
-            The predicted noise tensor for the guided diffusion process.
-        noise_pred_text (`torch.Tensor`):
-            The predicted noise tensor for the text-guided diffusion process.
-        guidance_rescale (`float`, *optional*, defaults to 0.0):
-            A rescale factor applied to the noise predictions.
-
-    Returns:
-        noise_cfg (`torch.Tensor`): The rescaled noise prediction tensor.
-    """
-    std_text = noise_pred_text.std(dim=list(range(1, noise_pred_text.ndim)), keepdim=True)
-    std_cfg = noise_cfg.std(dim=list(range(1, noise_cfg.ndim)), keepdim=True)
-    # rescale the results from guidance (fixes overexposure)
-    noise_pred_rescaled = noise_cfg * (std_text / std_cfg)
-    # mix with the original results from guidance by factor guidance_rescale to avoid "plain looking" images
-    noise_cfg = guidance_rescale * noise_pred_rescaled + (1 - guidance_rescale) * noise_cfg
-    return noise_cfg
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
 class StableDiffusionXLImg2ImgPipeline(
     DiffusionPipeline,
+    StableDiffusionXLMixin,
     StableDiffusionMixin,
     TextualInversionLoaderMixin,
     FromSingleFileMixin,
@@ -301,263 +198,6 @@ def __init__(
         else:
             self.watermark = None
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: str,
-        prompt_2: Optional[str] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        do_classifier_free_guidance: bool = True,
-        negative_prompt: Optional[str] = None,
-        negative_prompt_2: Optional[str] = None,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        negative_prompt_embeds: Optional[torch.Tensor] = None,
-        pooled_prompt_embeds: Optional[torch.Tensor] = None,
-        negative_pooled_prompt_embeds: Optional[torch.Tensor] = None,
-        lora_scale: Optional[float] = None,
-        clip_skip: Optional[int] = None,
-    ):
-        r"""
-        Encodes the prompt into text encoder hidden states.
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                used in both text-encoders
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            do_classifier_free_guidance (`bool`):
-                whether to use classifier free guidance or not
-            negative_prompt (`str` or `List[str]`, *optional*):
-                The prompt or prompts not to guide the image generation. If not defined, one has to pass
-                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
-                less than `1`).
-            negative_prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts not to guide the image generation to be sent to `tokenizer_2` and
-                `text_encoder_2`. If not defined, `negative_prompt` is used in both text-encoders
-            prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
-                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
-                argument.
-            pooled_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            negative_pooled_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated negative pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
-                weighting. If not provided, pooled negative_prompt_embeds will be generated from `negative_prompt`
-                input argument.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-            clip_skip (`int`, *optional*):
-                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
-                the output of the pre-final layer will be used for computing the prompt embeddings.
-        """
-        device = device or self._execution_device
-
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionXLLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None:
-                if not USE_PEFT_BACKEND:
-                    adjust_lora_scale_text_encoder(self.text_encoder, lora_scale)
-                else:
-                    scale_lora_layers(self.text_encoder, lora_scale)
-
-            if self.text_encoder_2 is not None:
-                if not USE_PEFT_BACKEND:
-                    adjust_lora_scale_text_encoder(self.text_encoder_2, lora_scale)
-                else:
-                    scale_lora_layers(self.text_encoder_2, lora_scale)
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        if prompt is not None:
-            batch_size = len(prompt)
-        else:
-            batch_size = prompt_embeds.shape[0]
-
-        # Define tokenizers and text encoders
-        tokenizers = [self.tokenizer, self.tokenizer_2] if self.tokenizer is not None else [self.tokenizer_2]
-        text_encoders = (
-            [self.text_encoder, self.text_encoder_2] if self.text_encoder is not None else [self.text_encoder_2]
-        )
-
-        if prompt_embeds is None:
-            prompt_2 = prompt_2 or prompt
-            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
-
-            # textual inversion: process multi-vector tokens if necessary
-            prompt_embeds_list = []
-            prompts = [prompt, prompt_2]
-            for prompt, tokenizer, text_encoder in zip(prompts, tokenizers, text_encoders):
-                if isinstance(self, TextualInversionLoaderMixin):
-                    prompt = self.maybe_convert_prompt(prompt, tokenizer)
-
-                text_inputs = tokenizer(
-                    prompt,
-                    padding="max_length",
-                    max_length=tokenizer.model_max_length,
-                    truncation=True,
-                    return_tensors="pt",
-                )
-
-                text_input_ids = text_inputs.input_ids
-                untruncated_ids = tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-
-                if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
-                    text_input_ids, untruncated_ids
-                ):
-                    removed_text = tokenizer.batch_decode(untruncated_ids[:, tokenizer.model_max_length - 1 : -1])
-                    logger.warning(
-                        "The following part of your input was truncated because CLIP can only handle sequences up to"
-                        f" {tokenizer.model_max_length} tokens: {removed_text}"
-                    )
-
-                prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)
-
-                # We are only ALWAYS interested in the pooled output of the final text encoder
-                if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
-                    pooled_prompt_embeds = prompt_embeds[0]
-
-                if clip_skip is None:
-                    prompt_embeds = prompt_embeds.hidden_states[-2]
-                else:
-                    # "2" because SDXL always indexes from the penultimate layer.
-                    prompt_embeds = prompt_embeds.hidden_states[-(clip_skip + 2)]
-
-                prompt_embeds_list.append(prompt_embeds)
-
-            prompt_embeds = torch.concat(prompt_embeds_list, dim=-1)
-
-        # get unconditional embeddings for classifier free guidance
-        zero_out_negative_prompt = negative_prompt is None and self.config.force_zeros_for_empty_prompt
-        if do_classifier_free_guidance and negative_prompt_embeds is None and zero_out_negative_prompt:
-            negative_prompt_embeds = torch.zeros_like(prompt_embeds)
-            negative_pooled_prompt_embeds = torch.zeros_like(pooled_prompt_embeds)
-        elif do_classifier_free_guidance and negative_prompt_embeds is None:
-            negative_prompt = negative_prompt or ""
-            negative_prompt_2 = negative_prompt_2 or negative_prompt
-
-            # normalize str to list
-            negative_prompt = batch_size * [negative_prompt] if isinstance(negative_prompt, str) else negative_prompt
-            negative_prompt_2 = (
-                batch_size * [negative_prompt_2] if isinstance(negative_prompt_2, str) else negative_prompt_2
-            )
-
-            uncond_tokens: List[str]
-            if prompt is not None and type(prompt) is not type(negative_prompt):
-                raise TypeError(
-                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
-                    f" {type(prompt)}."
-                )
-            elif batch_size != len(negative_prompt):
-                raise ValueError(
-                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
-                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
-                    " the batch size of `prompt`."
-                )
-            else:
-                uncond_tokens = [negative_prompt, negative_prompt_2]
-
-            negative_prompt_embeds_list = []
-            for negative_prompt, tokenizer, text_encoder in zip(uncond_tokens, tokenizers, text_encoders):
-                if isinstance(self, TextualInversionLoaderMixin):
-                    negative_prompt = self.maybe_convert_prompt(negative_prompt, tokenizer)
-
-                max_length = prompt_embeds.shape[1]
-                uncond_input = tokenizer(
-                    negative_prompt,
-                    padding="max_length",
-                    max_length=max_length,
-                    truncation=True,
-                    return_tensors="pt",
-                )
-
-                negative_prompt_embeds = text_encoder(
-                    uncond_input.input_ids.to(device),
-                    output_hidden_states=True,
-                )
-
-                # We are only ALWAYS interested in the pooled output of the final text encoder
-                if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
-                    negative_pooled_prompt_embeds = negative_prompt_embeds[0]
-                negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]
-
-                negative_prompt_embeds_list.append(negative_prompt_embeds)
-
-            negative_prompt_embeds = torch.concat(negative_prompt_embeds_list, dim=-1)
-
-        if self.text_encoder_2 is not None:
-            prompt_embeds = prompt_embeds.to(dtype=self.text_encoder_2.dtype, device=device)
-        else:
-            prompt_embeds = prompt_embeds.to(dtype=self.unet.dtype, device=device)
-
-        bs_embed, seq_len, _ = prompt_embeds.shape
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)
-
-        if do_classifier_free_guidance:
-            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
-            seq_len = negative_prompt_embeds.shape[1]
-
-            if self.text_encoder_2 is not None:
-                negative_prompt_embeds = negative_prompt_embeds.to(dtype=self.text_encoder_2.dtype, device=device)
-            else:
-                negative_prompt_embeds = negative_prompt_embeds.to(dtype=self.unet.dtype, device=device)
-
-            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
-            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        pooled_prompt_embeds = pooled_prompt_embeds.repeat(1, num_images_per_prompt).view(
-            bs_embed * num_images_per_prompt, -1
-        )
-        if do_classifier_free_guidance:
-            negative_pooled_prompt_embeds = negative_pooled_prompt_embeds.repeat(1, num_images_per_prompt).view(
-                bs_embed * num_images_per_prompt, -1
-            )
-
-        if self.text_encoder is not None:
-            if isinstance(self, StableDiffusionXLLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        if self.text_encoder_2 is not None:
-            if isinstance(self, StableDiffusionXLLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder_2, lora_scale)
-
-        return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
-    def prepare_extra_step_kwargs(self, generator, eta):
-        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
-        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
-        # eta corresponds to η in DDIM paper: https://huggingface.co/papers/2010.02502
-        # and should be between [0, 1]
-
-        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        extra_step_kwargs = {}
-        if accepts_eta:
-            extra_step_kwargs["eta"] = eta
-
-        # check if the scheduler accepts generator
-        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        if accepts_generator:
-            extra_step_kwargs["generator"] = generator
-        return extra_step_kwargs
-
     def check_inputs(
         self,
         prompt,
@@ -773,77 +413,6 @@ def prepare_latents(
 
         return latents
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_image
-    def encode_image(self, image, device, num_images_per_prompt, output_hidden_states=None):
-        dtype = next(self.image_encoder.parameters()).dtype
-
-        if not isinstance(image, torch.Tensor):
-            image = self.feature_extractor(image, return_tensors="pt").pixel_values
-
-        image = image.to(device=device, dtype=dtype)
-        if output_hidden_states:
-            image_enc_hidden_states = self.image_encoder(image, output_hidden_states=True).hidden_states[-2]
-            image_enc_hidden_states = image_enc_hidden_states.repeat_interleave(num_images_per_prompt, dim=0)
-            uncond_image_enc_hidden_states = self.image_encoder(
-                torch.zeros_like(image), output_hidden_states=True
-            ).hidden_states[-2]
-            uncond_image_enc_hidden_states = uncond_image_enc_hidden_states.repeat_interleave(
-                num_images_per_prompt, dim=0
-            )
-            return image_enc_hidden_states, uncond_image_enc_hidden_states
-        else:
-            image_embeds = self.image_encoder(image).image_embeds
-            image_embeds = image_embeds.repeat_interleave(num_images_per_prompt, dim=0)
-            uncond_image_embeds = torch.zeros_like(image_embeds)
-
-            return image_embeds, uncond_image_embeds
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_ip_adapter_image_embeds
-    def prepare_ip_adapter_image_embeds(
-        self, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt, do_classifier_free_guidance
-    ):
-        image_embeds = []
-        if do_classifier_free_guidance:
-            negative_image_embeds = []
-        if ip_adapter_image_embeds is None:
-            if not isinstance(ip_adapter_image, list):
-                ip_adapter_image = [ip_adapter_image]
-
-            if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
-                raise ValueError(
-                    f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
-                )
-
-            for single_ip_adapter_image, image_proj_layer in zip(
-                ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
-            ):
-                output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
-                single_image_embeds, single_negative_image_embeds = self.encode_image(
-                    single_ip_adapter_image, device, 1, output_hidden_state
-                )
-
-                image_embeds.append(single_image_embeds[None, :])
-                if do_classifier_free_guidance:
-                    negative_image_embeds.append(single_negative_image_embeds[None, :])
-        else:
-            for single_image_embeds in ip_adapter_image_embeds:
-                if do_classifier_free_guidance:
-                    single_negative_image_embeds, single_image_embeds = single_image_embeds.chunk(2)
-                    negative_image_embeds.append(single_negative_image_embeds)
-                image_embeds.append(single_image_embeds)
-
-        ip_adapter_image_embeds = []
-        for i, single_image_embeds in enumerate(image_embeds):
-            single_image_embeds = torch.cat([single_image_embeds] * num_images_per_prompt, dim=0)
-            if do_classifier_free_guidance:
-                single_negative_image_embeds = torch.cat([negative_image_embeds[i]] * num_images_per_prompt, dim=0)
-                single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds], dim=0)
-
-            single_image_embeds = single_image_embeds.to(device=device)
-            ip_adapter_image_embeds.append(single_image_embeds)
-
-        return ip_adapter_image_embeds
-
     def _get_add_time_ids(
         self,
         original_size,
@@ -895,94 +464,6 @@ def _get_add_time_ids(
 
         return add_time_ids, add_neg_time_ids
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_upscale.StableDiffusionUpscalePipeline.upcast_vae
-    def upcast_vae(self):
-        dtype = self.vae.dtype
-        self.vae.to(dtype=torch.float32)
-        use_torch_2_0_or_xformers = isinstance(
-            self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
-        )
-        # if xformers or torch_2_0 is used attention block does not need
-        # to be in float32 which can save lots of memory
-        if use_torch_2_0_or_xformers:
-            self.vae.post_quant_conv.to(dtype)
-            self.vae.decoder.conv_in.to(dtype)
-            self.vae.decoder.mid_block.to(dtype)
-
-    # Copied from diffusers.pipelines.latent_consistency_models.pipeline_latent_consistency_text2img.LatentConsistencyModelPipeline.get_guidance_scale_embedding
-    def get_guidance_scale_embedding(
-        self, w: torch.Tensor, embedding_dim: int = 512, dtype: torch.dtype = torch.float32
-    ) -> torch.Tensor:
-        """
-        See https://github.com/google-research/vdm/blob/dc27b98a554f65cdc654b800da5aa1846545d41b/model_vdm.py#L298
-
-        Args:
-            w (`torch.Tensor`):
-                Generate embedding vectors with a specified guidance scale to subsequently enrich timestep embeddings.
-            embedding_dim (`int`, *optional*, defaults to 512):
-                Dimension of the embeddings to generate.
-            dtype (`torch.dtype`, *optional*, defaults to `torch.float32`):
-                Data type of the generated embeddings.
-
-        Returns:
-            `torch.Tensor`: Embedding vectors with shape `(len(w), embedding_dim)`.
-        """
-        assert len(w.shape) == 1
-        w = w * 1000.0
-
-        half_dim = embedding_dim // 2
-        emb = torch.log(torch.tensor(10000.0)) / (half_dim - 1)
-        emb = torch.exp(torch.arange(half_dim, dtype=dtype) * -emb)
-        emb = w.to(dtype)[:, None] * emb[None, :]
-        emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=1)
-        if embedding_dim % 2 == 1:  # zero pad
-            emb = torch.nn.functional.pad(emb, (0, 1))
-        assert emb.shape == (w.shape[0], embedding_dim)
-        return emb
-
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def guidance_rescale(self):
-        return self._guidance_rescale
-
-    @property
-    def clip_skip(self):
-        return self._clip_skip
-
-    # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
-    # of the Imagen paper: https://huggingface.co/papers/2205.11487 . `guidance_scale = 1`
-    # corresponds to doing no classifier free guidance.
-    @property
-    def do_classifier_free_guidance(self):
-        return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
-
-    @property
-    def cross_attention_kwargs(self):
-        return self._cross_attention_kwargs
-
-    @property
-    def denoising_end(self):
-        return self._denoising_end
-
-    @property
-    def denoising_start(self):
-        return self._denoising_start
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_inpaint.py b/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_inpaint.py
index 88cc7515b051..6c6925bba0f0 100644
--- a/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_inpaint.py
+++ b/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_inpaint.py
@@ -12,10 +12,8 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Tuple, Union
 
-import numpy as np
 import PIL.Image
 import torch
 from transformers import (
@@ -34,26 +32,24 @@
     StableDiffusionXLLoraLoaderMixin,
     TextualInversionLoaderMixin,
 )
-from ...models import AutoencoderKL, ImageProjection, UNet2DConditionModel
-from ...models.attention_processor import (
-    AttnProcessor2_0,
-    XFormersAttnProcessor,
-)
-from ...models.lora import adjust_lora_scale_text_encoder
+from ...models import AutoencoderKL, UNet2DConditionModel
 from ...schedulers import KarrasDiffusionSchedulers
 from ...utils import (
-    USE_PEFT_BACKEND,
     deprecate,
     is_invisible_watermark_available,
     is_torch_xla_available,
     logging,
     replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
 )
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from .pipeline_output import StableDiffusionXLPipelineOutput
+from .pipeline_stable_diffusion_xl_utils import (
+    StableDiffusionXLMixin,
+    rescale_noise_cfg,
+    retrieve_latents,
+    retrieve_timesteps,
+)
 
 
 if is_invisible_watermark_available():
@@ -99,126 +95,10 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
-def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
-    r"""
-    Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
-    Section 3.4 from [Common Diffusion Noise Schedules and Sample Steps are
-    Flawed](https://huggingface.co/papers/2305.08891).
-
-    Args:
-        noise_cfg (`torch.Tensor`):
-            The predicted noise tensor for the guided diffusion process.
-        noise_pred_text (`torch.Tensor`):
-            The predicted noise tensor for the text-guided diffusion process.
-        guidance_rescale (`float`, *optional*, defaults to 0.0):
-            A rescale factor applied to the noise predictions.
-
-    Returns:
-        noise_cfg (`torch.Tensor`): The rescaled noise prediction tensor.
-    """
-    std_text = noise_pred_text.std(dim=list(range(1, noise_pred_text.ndim)), keepdim=True)
-    std_cfg = noise_cfg.std(dim=list(range(1, noise_cfg.ndim)), keepdim=True)
-    # rescale the results from guidance (fixes overexposure)
-    noise_pred_rescaled = noise_cfg * (std_text / std_cfg)
-    # mix with the original results from guidance by factor guidance_rescale to avoid "plain looking" images
-    noise_cfg = guidance_rescale * noise_pred_rescaled + (1 - guidance_rescale) * noise_cfg
-    return noise_cfg
-
-
-def mask_pil_to_torch(mask, height, width):
-    # preprocess mask
-    if isinstance(mask, (PIL.Image.Image, np.ndarray)):
-        mask = [mask]
-
-    if isinstance(mask, list) and isinstance(mask[0], PIL.Image.Image):
-        mask = [i.resize((width, height), resample=PIL.Image.LANCZOS) for i in mask]
-        mask = np.concatenate([np.array(m.convert("L"))[None, None, :] for m in mask], axis=0)
-        mask = mask.astype(np.float32) / 255.0
-    elif isinstance(mask, list) and isinstance(mask[0], np.ndarray):
-        mask = np.concatenate([m[None, None, :] for m in mask], axis=0)
-
-    mask = torch.from_numpy(mask)
-    return mask
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    r"""
-    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
-    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
-
-    Args:
-        scheduler (`SchedulerMixin`):
-            The scheduler to get timesteps from.
-        num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
-            must be `None`.
-        device (`str` or `torch.device`, *optional*):
-            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
-        timesteps (`List[int]`, *optional*):
-            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
-            `num_inference_steps` and `sigmas` must be `None`.
-        sigmas (`List[float]`, *optional*):
-            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
-            `num_inference_steps` and `timesteps` must be `None`.
-
-    Returns:
-        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
-        second element is the number of inference steps.
-    """
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accepts_timesteps:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" timestep schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
-        if not accept_sigmas:
-            raise ValueError(
-                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
-                f" sigmas schedules. Please check whether you are using the correct scheduler."
-            )
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-
 class StableDiffusionXLInpaintPipeline(
     DiffusionPipeline,
     StableDiffusionMixin,
+    StableDiffusionXLMixin,
     TextualInversionLoaderMixin,
     StableDiffusionXLLoraLoaderMixin,
     FromSingleFileMixin,
@@ -334,334 +214,6 @@ def __init__(
         else:
             self.watermark = None
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_image
-    def encode_image(self, image, device, num_images_per_prompt, output_hidden_states=None):
-        dtype = next(self.image_encoder.parameters()).dtype
-
-        if not isinstance(image, torch.Tensor):
-            image = self.feature_extractor(image, return_tensors="pt").pixel_values
-
-        image = image.to(device=device, dtype=dtype)
-        if output_hidden_states:
-            image_enc_hidden_states = self.image_encoder(image, output_hidden_states=True).hidden_states[-2]
-            image_enc_hidden_states = image_enc_hidden_states.repeat_interleave(num_images_per_prompt, dim=0)
-            uncond_image_enc_hidden_states = self.image_encoder(
-                torch.zeros_like(image), output_hidden_states=True
-            ).hidden_states[-2]
-            uncond_image_enc_hidden_states = uncond_image_enc_hidden_states.repeat_interleave(
-                num_images_per_prompt, dim=0
-            )
-            return image_enc_hidden_states, uncond_image_enc_hidden_states
-        else:
-            image_embeds = self.image_encoder(image).image_embeds
-            image_embeds = image_embeds.repeat_interleave(num_images_per_prompt, dim=0)
-            uncond_image_embeds = torch.zeros_like(image_embeds)
-
-            return image_embeds, uncond_image_embeds
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_ip_adapter_image_embeds
-    def prepare_ip_adapter_image_embeds(
-        self, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt, do_classifier_free_guidance
-    ):
-        image_embeds = []
-        if do_classifier_free_guidance:
-            negative_image_embeds = []
-        if ip_adapter_image_embeds is None:
-            if not isinstance(ip_adapter_image, list):
-                ip_adapter_image = [ip_adapter_image]
-
-            if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
-                raise ValueError(
-                    f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
-                )
-
-            for single_ip_adapter_image, image_proj_layer in zip(
-                ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
-            ):
-                output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
-                single_image_embeds, single_negative_image_embeds = self.encode_image(
-                    single_ip_adapter_image, device, 1, output_hidden_state
-                )
-
-                image_embeds.append(single_image_embeds[None, :])
-                if do_classifier_free_guidance:
-                    negative_image_embeds.append(single_negative_image_embeds[None, :])
-        else:
-            for single_image_embeds in ip_adapter_image_embeds:
-                if do_classifier_free_guidance:
-                    single_negative_image_embeds, single_image_embeds = single_image_embeds.chunk(2)
-                    negative_image_embeds.append(single_negative_image_embeds)
-                image_embeds.append(single_image_embeds)
-
-        ip_adapter_image_embeds = []
-        for i, single_image_embeds in enumerate(image_embeds):
-            single_image_embeds = torch.cat([single_image_embeds] * num_images_per_prompt, dim=0)
-            if do_classifier_free_guidance:
-                single_negative_image_embeds = torch.cat([negative_image_embeds[i]] * num_images_per_prompt, dim=0)
-                single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds], dim=0)
-
-            single_image_embeds = single_image_embeds.to(device=device)
-            ip_adapter_image_embeds.append(single_image_embeds)
-
-        return ip_adapter_image_embeds
-
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
-    def encode_prompt(
-        self,
-        prompt: str,
-        prompt_2: Optional[str] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        do_classifier_free_guidance: bool = True,
-        negative_prompt: Optional[str] = None,
-        negative_prompt_2: Optional[str] = None,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        negative_prompt_embeds: Optional[torch.Tensor] = None,
-        pooled_prompt_embeds: Optional[torch.Tensor] = None,
-        negative_pooled_prompt_embeds: Optional[torch.Tensor] = None,
-        lora_scale: Optional[float] = None,
-        clip_skip: Optional[int] = None,
-    ):
-        r"""
-        Encodes the prompt into text encoder hidden states.
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                used in both text-encoders
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            do_classifier_free_guidance (`bool`):
-                whether to use classifier free guidance or not
-            negative_prompt (`str` or `List[str]`, *optional*):
-                The prompt or prompts not to guide the image generation. If not defined, one has to pass
-                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
-                less than `1`).
-            negative_prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts not to guide the image generation to be sent to `tokenizer_2` and
-                `text_encoder_2`. If not defined, `negative_prompt` is used in both text-encoders
-            prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
-                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
-                argument.
-            pooled_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            negative_pooled_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated negative pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
-                weighting. If not provided, pooled negative_prompt_embeds will be generated from `negative_prompt`
-                input argument.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-            clip_skip (`int`, *optional*):
-                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
-                the output of the pre-final layer will be used for computing the prompt embeddings.
-        """
-        device = device or self._execution_device
-
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionXLLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None:
-                if not USE_PEFT_BACKEND:
-                    adjust_lora_scale_text_encoder(self.text_encoder, lora_scale)
-                else:
-                    scale_lora_layers(self.text_encoder, lora_scale)
-
-            if self.text_encoder_2 is not None:
-                if not USE_PEFT_BACKEND:
-                    adjust_lora_scale_text_encoder(self.text_encoder_2, lora_scale)
-                else:
-                    scale_lora_layers(self.text_encoder_2, lora_scale)
-
-        prompt = [prompt] if isinstance(prompt, str) else prompt
-
-        if prompt is not None:
-            batch_size = len(prompt)
-        else:
-            batch_size = prompt_embeds.shape[0]
-
-        # Define tokenizers and text encoders
-        tokenizers = [self.tokenizer, self.tokenizer_2] if self.tokenizer is not None else [self.tokenizer_2]
-        text_encoders = (
-            [self.text_encoder, self.text_encoder_2] if self.text_encoder is not None else [self.text_encoder_2]
-        )
-
-        if prompt_embeds is None:
-            prompt_2 = prompt_2 or prompt
-            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
-
-            # textual inversion: process multi-vector tokens if necessary
-            prompt_embeds_list = []
-            prompts = [prompt, prompt_2]
-            for prompt, tokenizer, text_encoder in zip(prompts, tokenizers, text_encoders):
-                if isinstance(self, TextualInversionLoaderMixin):
-                    prompt = self.maybe_convert_prompt(prompt, tokenizer)
-
-                text_inputs = tokenizer(
-                    prompt,
-                    padding="max_length",
-                    max_length=tokenizer.model_max_length,
-                    truncation=True,
-                    return_tensors="pt",
-                )
-
-                text_input_ids = text_inputs.input_ids
-                untruncated_ids = tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-
-                if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
-                    text_input_ids, untruncated_ids
-                ):
-                    removed_text = tokenizer.batch_decode(untruncated_ids[:, tokenizer.model_max_length - 1 : -1])
-                    logger.warning(
-                        "The following part of your input was truncated because CLIP can only handle sequences up to"
-                        f" {tokenizer.model_max_length} tokens: {removed_text}"
-                    )
-
-                prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)
-
-                # We are only ALWAYS interested in the pooled output of the final text encoder
-                if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
-                    pooled_prompt_embeds = prompt_embeds[0]
-
-                if clip_skip is None:
-                    prompt_embeds = prompt_embeds.hidden_states[-2]
-                else:
-                    # "2" because SDXL always indexes from the penultimate layer.
-                    prompt_embeds = prompt_embeds.hidden_states[-(clip_skip + 2)]
-
-                prompt_embeds_list.append(prompt_embeds)
-
-            prompt_embeds = torch.concat(prompt_embeds_list, dim=-1)
-
-        # get unconditional embeddings for classifier free guidance
-        zero_out_negative_prompt = negative_prompt is None and self.config.force_zeros_for_empty_prompt
-        if do_classifier_free_guidance and negative_prompt_embeds is None and zero_out_negative_prompt:
-            negative_prompt_embeds = torch.zeros_like(prompt_embeds)
-            negative_pooled_prompt_embeds = torch.zeros_like(pooled_prompt_embeds)
-        elif do_classifier_free_guidance and negative_prompt_embeds is None:
-            negative_prompt = negative_prompt or ""
-            negative_prompt_2 = negative_prompt_2 or negative_prompt
-
-            # normalize str to list
-            negative_prompt = batch_size * [negative_prompt] if isinstance(negative_prompt, str) else negative_prompt
-            negative_prompt_2 = (
-                batch_size * [negative_prompt_2] if isinstance(negative_prompt_2, str) else negative_prompt_2
-            )
-
-            uncond_tokens: List[str]
-            if prompt is not None and type(prompt) is not type(negative_prompt):
-                raise TypeError(
-                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
-                    f" {type(prompt)}."
-                )
-            elif batch_size != len(negative_prompt):
-                raise ValueError(
-                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
-                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
-                    " the batch size of `prompt`."
-                )
-            else:
-                uncond_tokens = [negative_prompt, negative_prompt_2]
-
-            negative_prompt_embeds_list = []
-            for negative_prompt, tokenizer, text_encoder in zip(uncond_tokens, tokenizers, text_encoders):
-                if isinstance(self, TextualInversionLoaderMixin):
-                    negative_prompt = self.maybe_convert_prompt(negative_prompt, tokenizer)
-
-                max_length = prompt_embeds.shape[1]
-                uncond_input = tokenizer(
-                    negative_prompt,
-                    padding="max_length",
-                    max_length=max_length,
-                    truncation=True,
-                    return_tensors="pt",
-                )
-
-                negative_prompt_embeds = text_encoder(
-                    uncond_input.input_ids.to(device),
-                    output_hidden_states=True,
-                )
-
-                # We are only ALWAYS interested in the pooled output of the final text encoder
-                if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
-                    negative_pooled_prompt_embeds = negative_prompt_embeds[0]
-                negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]
-
-                negative_prompt_embeds_list.append(negative_prompt_embeds)
-
-            negative_prompt_embeds = torch.concat(negative_prompt_embeds_list, dim=-1)
-
-        if self.text_encoder_2 is not None:
-            prompt_embeds = prompt_embeds.to(dtype=self.text_encoder_2.dtype, device=device)
-        else:
-            prompt_embeds = prompt_embeds.to(dtype=self.unet.dtype, device=device)
-
-        bs_embed, seq_len, _ = prompt_embeds.shape
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)
-
-        if do_classifier_free_guidance:
-            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
-            seq_len = negative_prompt_embeds.shape[1]
-
-            if self.text_encoder_2 is not None:
-                negative_prompt_embeds = negative_prompt_embeds.to(dtype=self.text_encoder_2.dtype, device=device)
-            else:
-                negative_prompt_embeds = negative_prompt_embeds.to(dtype=self.unet.dtype, device=device)
-
-            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
-            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        pooled_prompt_embeds = pooled_prompt_embeds.repeat(1, num_images_per_prompt).view(
-            bs_embed * num_images_per_prompt, -1
-        )
-        if do_classifier_free_guidance:
-            negative_pooled_prompt_embeds = negative_pooled_prompt_embeds.repeat(1, num_images_per_prompt).view(
-                bs_embed * num_images_per_prompt, -1
-            )
-
-        if self.text_encoder is not None:
-            if isinstance(self, StableDiffusionXLLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        if self.text_encoder_2 is not None:
-            if isinstance(self, StableDiffusionXLLoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder_2, lora_scale)
-
-        return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
-    def prepare_extra_step_kwargs(self, generator, eta):
-        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
-        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
-        # eta corresponds to η in DDIM paper: https://huggingface.co/papers/2010.02502
-        # and should be between [0, 1]
-
-        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        extra_step_kwargs = {}
-        if accepts_eta:
-            extra_step_kwargs["eta"] = eta
-
-        # check if the scheduler accepts generator
-        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        if accepts_generator:
-            extra_step_kwargs["generator"] = generator
-        return extra_step_kwargs
-
     def check_inputs(
         self,
         prompt,
@@ -1000,94 +552,6 @@ def _get_add_time_ids(
 
         return add_time_ids, add_neg_time_ids
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_upscale.StableDiffusionUpscalePipeline.upcast_vae
-    def upcast_vae(self):
-        dtype = self.vae.dtype
-        self.vae.to(dtype=torch.float32)
-        use_torch_2_0_or_xformers = isinstance(
-            self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
-        )
-        # if xformers or torch_2_0 is used attention block does not need
-        # to be in float32 which can save lots of memory
-        if use_torch_2_0_or_xformers:
-            self.vae.post_quant_conv.to(dtype)
-            self.vae.decoder.conv_in.to(dtype)
-            self.vae.decoder.mid_block.to(dtype)
-
-    # Copied from diffusers.pipelines.latent_consistency_models.pipeline_latent_consistency_text2img.LatentConsistencyModelPipeline.get_guidance_scale_embedding
-    def get_guidance_scale_embedding(
-        self, w: torch.Tensor, embedding_dim: int = 512, dtype: torch.dtype = torch.float32
-    ) -> torch.Tensor:
-        """
-        See https://github.com/google-research/vdm/blob/dc27b98a554f65cdc654b800da5aa1846545d41b/model_vdm.py#L298
-
-        Args:
-            w (`torch.Tensor`):
-                Generate embedding vectors with a specified guidance scale to subsequently enrich timestep embeddings.
-            embedding_dim (`int`, *optional*, defaults to 512):
-                Dimension of the embeddings to generate.
-            dtype (`torch.dtype`, *optional*, defaults to `torch.float32`):
-                Data type of the generated embeddings.
-
-        Returns:
-            `torch.Tensor`: Embedding vectors with shape `(len(w), embedding_dim)`.
-        """
-        assert len(w.shape) == 1
-        w = w * 1000.0
-
-        half_dim = embedding_dim // 2
-        emb = torch.log(torch.tensor(10000.0)) / (half_dim - 1)
-        emb = torch.exp(torch.arange(half_dim, dtype=dtype) * -emb)
-        emb = w.to(dtype)[:, None] * emb[None, :]
-        emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=1)
-        if embedding_dim % 2 == 1:  # zero pad
-            emb = torch.nn.functional.pad(emb, (0, 1))
-        assert emb.shape == (w.shape[0], embedding_dim)
-        return emb
-
-    @property
-    def guidance_scale(self):
-        return self._guidance_scale
-
-    @property
-    def guidance_rescale(self):
-        return self._guidance_rescale
-
-    @property
-    def clip_skip(self):
-        return self._clip_skip
-
-    # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
-    # of the Imagen paper: https://huggingface.co/papers/2205.11487 . `guidance_scale = 1`
-    # corresponds to doing no classifier free guidance.
-    @property
-    def do_classifier_free_guidance(self):
-        return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
-
-    @property
-    def cross_attention_kwargs(self):
-        return self._cross_attention_kwargs
-
-    @property
-    def denoising_end(self):
-        return self._denoising_end
-
-    @property
-    def denoising_start(self):
-        return self._denoising_start
-
-    @property
-    def num_timesteps(self):
-        return self._num_timesteps
-
-    @property
-    def interrupt(self):
-        return self._interrupt
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_instruct_pix2pix.py b/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_instruct_pix2pix.py
index 58b008361782..d3033a4cc6e9 100644
--- a/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_instruct_pix2pix.py
+++ b/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_instruct_pix2pix.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import inspect
 from typing import Any, Callable, Dict, List, Optional, Tuple, Union
 
 import PIL.Image
@@ -22,25 +21,18 @@
 from ...image_processor import PipelineImageInput, VaeImageProcessor
 from ...loaders import FromSingleFileMixin, StableDiffusionXLLoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, UNet2DConditionModel
-from ...models.attention_processor import (
-    AttnProcessor2_0,
-    FusedAttnProcessor2_0,
-    XFormersAttnProcessor,
-)
-from ...models.lora import adjust_lora_scale_text_encoder
 from ...schedulers import KarrasDiffusionSchedulers
 from ...utils import (
-    USE_PEFT_BACKEND,
     deprecate,
     is_invisible_watermark_available,
     is_torch_xla_available,
     logging,
     replace_example_docstring,
-    scale_lora_layers,
 )
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from .pipeline_output import StableDiffusionXLPipelineOutput
+from .pipeline_stable_diffusion_xl_utils import StableDiffusionXLMixin, rescale_noise_cfg, retrieve_latents
 
 
 if is_invisible_watermark_available():
@@ -87,36 +79,9 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(
-    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
-):
-    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
-        return encoder_output.latent_dist.sample(generator)
-    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
-        return encoder_output.latent_dist.mode()
-    elif hasattr(encoder_output, "latents"):
-        return encoder_output.latents
-    else:
-        raise AttributeError("Could not access latents of provided encoder_output")
-
-
-def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
-    """
-    Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and
-    Sample Steps are Flawed](https://huggingface.co/papers/2305.08891). See Section 3.4
-    """
-    std_text = noise_pred_text.std(dim=list(range(1, noise_pred_text.ndim)), keepdim=True)
-    std_cfg = noise_cfg.std(dim=list(range(1, noise_cfg.ndim)), keepdim=True)
-    # rescale the results from guidance (fixes overexposure)
-    noise_pred_rescaled = noise_cfg * (std_text / std_cfg)
-    # mix with the original results from guidance by factor guidance_rescale to avoid "plain looking" images
-    noise_cfg = guidance_rescale * noise_pred_rescaled + (1 - guidance_rescale) * noise_cfg
-    return noise_cfg
-
-
 class StableDiffusionXLInstructPix2PixPipeline(
     DiffusionPipeline,
+    StableDiffusionXLMixin,
     StableDiffusionMixin,
     TextualInversionLoaderMixin,
     FromSingleFileMixin,
@@ -215,232 +180,6 @@ def __init__(
         else:
             self.watermark = None
 
-    def encode_prompt(
-        self,
-        prompt: str,
-        prompt_2: Optional[str] = None,
-        device: Optional[torch.device] = None,
-        num_images_per_prompt: int = 1,
-        do_classifier_free_guidance: bool = True,
-        negative_prompt: Optional[str] = None,
-        negative_prompt_2: Optional[str] = None,
-        prompt_embeds: Optional[torch.Tensor] = None,
-        negative_prompt_embeds: Optional[torch.Tensor] = None,
-        pooled_prompt_embeds: Optional[torch.Tensor] = None,
-        negative_pooled_prompt_embeds: Optional[torch.Tensor] = None,
-        lora_scale: Optional[float] = None,
-    ):
-        r"""
-        Encodes the prompt into text encoder hidden states.
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                prompt to be encoded
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                used in both text-encoders
-            device: (`torch.device`):
-                torch device
-            num_images_per_prompt (`int`):
-                number of images that should be generated per prompt
-            do_classifier_free_guidance (`bool`):
-                whether to use classifier free guidance or not
-            negative_prompt (`str` or `List[str]`, *optional*):
-                The prompt or prompts not to guide the image generation. If not defined, one has to pass
-                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
-                less than `1`).
-            negative_prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts not to guide the image generation to be sent to `tokenizer_2` and
-                `text_encoder_2`. If not defined, `negative_prompt` is used in both text-encoders
-            prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
-                provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
-                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
-                argument.
-            pooled_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            negative_pooled_prompt_embeds (`torch.Tensor`, *optional*):
-                Pre-generated negative pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
-                weighting. If not provided, pooled negative_prompt_embeds will be generated from `negative_prompt`
-                input argument.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
-        """
-        device = device or self._execution_device
-
-        # set lora scale so that monkey patched LoRA
-        # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionXLLoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None:
-                if not USE_PEFT_BACKEND:
-                    adjust_lora_scale_text_encoder(self.text_encoder, lora_scale)
-                else:
-                    scale_lora_layers(self.text_encoder, lora_scale)
-
-            if self.text_encoder_2 is not None:
-                if not USE_PEFT_BACKEND:
-                    adjust_lora_scale_text_encoder(self.text_encoder_2, lora_scale)
-                else:
-                    scale_lora_layers(self.text_encoder_2, lora_scale)
-
-        if prompt is not None and isinstance(prompt, str):
-            batch_size = 1
-        elif prompt is not None and isinstance(prompt, list):
-            batch_size = len(prompt)
-        else:
-            batch_size = prompt_embeds.shape[0]
-
-        # Define tokenizers and text encoders
-        tokenizers = [self.tokenizer, self.tokenizer_2] if self.tokenizer is not None else [self.tokenizer_2]
-        text_encoders = (
-            [self.text_encoder, self.text_encoder_2] if self.text_encoder is not None else [self.text_encoder_2]
-        )
-
-        if prompt_embeds is None:
-            prompt_2 = prompt_2 or prompt
-            # textual inversion: process multi-vector tokens if necessary
-            prompt_embeds_list = []
-            prompts = [prompt, prompt_2]
-            for prompt, tokenizer, text_encoder in zip(prompts, tokenizers, text_encoders):
-                if isinstance(self, TextualInversionLoaderMixin):
-                    prompt = self.maybe_convert_prompt(prompt, tokenizer)
-
-                text_inputs = tokenizer(
-                    prompt,
-                    padding="max_length",
-                    max_length=tokenizer.model_max_length,
-                    truncation=True,
-                    return_tensors="pt",
-                )
-
-                text_input_ids = text_inputs.input_ids
-                untruncated_ids = tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
-
-                if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
-                    text_input_ids, untruncated_ids
-                ):
-                    removed_text = tokenizer.batch_decode(untruncated_ids[:, tokenizer.model_max_length - 1 : -1])
-                    logger.warning(
-                        "The following part of your input was truncated because CLIP can only handle sequences up to"
-                        f" {tokenizer.model_max_length} tokens: {removed_text}"
-                    )
-
-                prompt_embeds = text_encoder(
-                    text_input_ids.to(device),
-                    output_hidden_states=True,
-                )
-
-                # We are only ALWAYS interested in the pooled output of the final text encoder
-                if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
-                    pooled_prompt_embeds = prompt_embeds[0]
-
-                prompt_embeds = prompt_embeds.hidden_states[-2]
-
-                prompt_embeds_list.append(prompt_embeds)
-
-            prompt_embeds = torch.concat(prompt_embeds_list, dim=-1)
-
-        # get unconditional embeddings for classifier free guidance
-        zero_out_negative_prompt = negative_prompt is None and self.config.force_zeros_for_empty_prompt
-        if do_classifier_free_guidance and negative_prompt_embeds is None and zero_out_negative_prompt:
-            negative_prompt_embeds = torch.zeros_like(prompt_embeds)
-            negative_pooled_prompt_embeds = torch.zeros_like(pooled_prompt_embeds)
-        elif do_classifier_free_guidance and negative_prompt_embeds is None:
-            negative_prompt = negative_prompt or ""
-            negative_prompt_2 = negative_prompt_2 or negative_prompt
-
-            uncond_tokens: List[str]
-            if prompt is not None and type(prompt) is not type(negative_prompt):
-                raise TypeError(
-                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
-                    f" {type(prompt)}."
-                )
-            elif isinstance(negative_prompt, str):
-                uncond_tokens = [negative_prompt, negative_prompt_2]
-            elif batch_size != len(negative_prompt):
-                raise ValueError(
-                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
-                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
-                    " the batch size of `prompt`."
-                )
-            else:
-                uncond_tokens = [negative_prompt, negative_prompt_2]
-
-            negative_prompt_embeds_list = []
-            for negative_prompt, tokenizer, text_encoder in zip(uncond_tokens, tokenizers, text_encoders):
-                if isinstance(self, TextualInversionLoaderMixin):
-                    negative_prompt = self.maybe_convert_prompt(negative_prompt, tokenizer)
-
-                max_length = prompt_embeds.shape[1]
-                uncond_input = tokenizer(
-                    negative_prompt,
-                    padding="max_length",
-                    max_length=max_length,
-                    truncation=True,
-                    return_tensors="pt",
-                )
-
-                negative_prompt_embeds = text_encoder(
-                    uncond_input.input_ids.to(device),
-                    output_hidden_states=True,
-                )
-                # We are only ALWAYS interested in the pooled output of the final text encoder
-                if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
-                    negative_pooled_prompt_embeds = negative_prompt_embeds[0]
-                negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]
-
-                negative_prompt_embeds_list.append(negative_prompt_embeds)
-
-            negative_prompt_embeds = torch.concat(negative_prompt_embeds_list, dim=-1)
-
-        prompt_embeds_dtype = self.text_encoder_2.dtype if self.text_encoder_2 is not None else self.unet.dtype
-        prompt_embeds = prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
-        bs_embed, seq_len, _ = prompt_embeds.shape
-        # duplicate text embeddings for each generation per prompt, using mps friendly method
-        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
-        prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)
-
-        if do_classifier_free_guidance:
-            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
-            seq_len = negative_prompt_embeds.shape[1]
-            negative_prompt_embeds = negative_prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
-            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
-            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
-
-        pooled_prompt_embeds = pooled_prompt_embeds.repeat(1, num_images_per_prompt).view(
-            bs_embed * num_images_per_prompt, -1
-        )
-        if do_classifier_free_guidance:
-            negative_pooled_prompt_embeds = negative_pooled_prompt_embeds.repeat(1, num_images_per_prompt).view(
-                bs_embed * num_images_per_prompt, -1
-            )
-
-        return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
-    def prepare_extra_step_kwargs(self, generator, eta):
-        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
-        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
-        # eta corresponds to η in DDIM paper: https://huggingface.co/papers/2010.02502
-        # and should be between [0, 1]
-
-        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        extra_step_kwargs = {}
-        if accepts_eta:
-            extra_step_kwargs["eta"] = eta
-
-        # check if the scheduler accepts generator
-        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
-        if accepts_generator:
-            extra_step_kwargs["generator"] = generator
-        return extra_step_kwargs
-
     def check_inputs(
         self,
         prompt,
@@ -588,25 +327,6 @@ def _get_add_time_ids(
         add_time_ids = torch.tensor([add_time_ids], dtype=dtype)
         return add_time_ids
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.upcast_vae
-    def upcast_vae(self):
-        dtype = self.vae.dtype
-        self.vae.to(dtype=torch.float32)
-        use_torch_2_0_or_xformers = isinstance(
-            self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-                FusedAttnProcessor2_0,
-            ),
-        )
-        # if xformers or torch_2_0 is used attention block does not need
-        # to be in float32 which can save lots of memory
-        if use_torch_2_0_or_xformers:
-            self.vae.post_quant_conv.to(dtype)
-            self.vae.decoder.conv_in.to(dtype)
-            self.vae.decoder.mid_block.to(dtype)
-
     @torch.no_grad()
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
diff --git a/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_utils.py b/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_utils.py
new file mode 100644
index 000000000000..c491b8d0bb4a
--- /dev/null
+++ b/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_utils.py
@@ -0,0 +1,541 @@
+# Copyright 2025 The Stability AI and HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import inspect
+from typing import List, Optional, Union
+
+import torch
+
+from ...loaders import StableDiffusionXLLoraLoaderMixin, TextualInversionLoaderMixin
+from ...models import ImageProjection
+from ...models.attention_processor import AttnProcessor2_0, FusedAttnProcessor2_0, XFormersAttnProcessor
+from ...models.lora import adjust_lora_scale_text_encoder
+from ...utils import USE_PEFT_BACKEND, logging, scale_lora_layers, unscale_lora_layers
+
+
+logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
+
+
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
+def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
+    r"""
+    Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
+    Section 3.4 from [Common Diffusion Noise Schedules and Sample Steps are
+    Flawed](https://huggingface.co/papers/2305.08891).
+
+    Args:
+        noise_cfg (`torch.Tensor`):
+            The predicted noise tensor for the guided diffusion process.
+        noise_pred_text (`torch.Tensor`):
+            The predicted noise tensor for the text-guided diffusion process.
+        guidance_rescale (`float`, *optional*, defaults to 0.0):
+            A rescale factor applied to the noise predictions.
+
+    Returns:
+        noise_cfg (`torch.Tensor`): The rescaled noise prediction tensor.
+    """
+    std_text = noise_pred_text.std(dim=list(range(1, noise_pred_text.ndim)), keepdim=True)
+    std_cfg = noise_cfg.std(dim=list(range(1, noise_cfg.ndim)), keepdim=True)
+    # rescale the results from guidance (fixes overexposure)
+    noise_pred_rescaled = noise_cfg * (std_text / std_cfg)
+    # mix with the original results from guidance by factor guidance_rescale to avoid "plain looking" images
+    noise_cfg = guidance_rescale * noise_pred_rescaled + (1 - guidance_rescale) * noise_cfg
+    return noise_cfg
+
+
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
+def retrieve_timesteps(
+    scheduler,
+    num_inference_steps: Optional[int] = None,
+    device: Optional[Union[str, torch.device]] = None,
+    timesteps: Optional[List[int]] = None,
+    sigmas: Optional[List[float]] = None,
+    **kwargs,
+):
+    r"""
+    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
+    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
+
+    Args:
+        scheduler (`SchedulerMixin`):
+            The scheduler to get timesteps from.
+        num_inference_steps (`int`):
+            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
+            must be `None`.
+        device (`str` or `torch.device`, *optional*):
+            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
+        timesteps (`List[int]`, *optional*):
+            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
+            `num_inference_steps` and `sigmas` must be `None`.
+        sigmas (`List[float]`, *optional*):
+            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
+            `num_inference_steps` and `timesteps` must be `None`.
+
+    Returns:
+        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
+        second element is the number of inference steps.
+    """
+    if timesteps is not None and sigmas is not None:
+        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
+    if timesteps is not None:
+        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accepts_timesteps:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" timestep schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    elif sigmas is not None:
+        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accept_sigmas:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" sigmas schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    else:
+        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+    return timesteps, num_inference_steps
+
+
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
+def retrieve_latents(
+    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
+):
+    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
+        return encoder_output.latent_dist.sample(generator)
+    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
+        return encoder_output.latent_dist.mode()
+    elif hasattr(encoder_output, "latents"):
+        return encoder_output.latents
+    else:
+        raise AttributeError("Could not access latents of provided encoder_output")
+
+
+class StableDiffusionXLMixin:
+    def encode_prompt(
+        self,
+        prompt: str,
+        prompt_2: Optional[str] = None,
+        device: Optional[torch.device] = None,
+        num_images_per_prompt: int = 1,
+        do_classifier_free_guidance: bool = True,
+        negative_prompt: Optional[str] = None,
+        negative_prompt_2: Optional[str] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
+        pooled_prompt_embeds: Optional[torch.Tensor] = None,
+        negative_pooled_prompt_embeds: Optional[torch.Tensor] = None,
+        lora_scale: Optional[float] = None,
+        clip_skip: Optional[int] = None,
+    ):
+        r"""
+        Encodes the prompt into text encoder hidden states.
+
+        Args:
+            prompt (`str` or `List[str]`, *optional*):
+                prompt to be encoded
+            prompt_2 (`str` or `List[str]`, *optional*):
+                The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
+                used in both text-encoders
+            device: (`torch.device`):
+                torch device
+            num_images_per_prompt (`int`):
+                number of images that should be generated per prompt
+            do_classifier_free_guidance (`bool`):
+                whether to use classifier free guidance or not
+            negative_prompt (`str` or `List[str]`, *optional*):
+                The prompt or prompts not to guide the image generation. If not defined, one has to pass
+                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
+                less than `1`).
+            negative_prompt_2 (`str` or `List[str]`, *optional*):
+                The prompt or prompts not to guide the image generation to be sent to `tokenizer_2` and
+                `text_encoder_2`. If not defined, `negative_prompt` is used in both text-encoders
+            prompt_embeds (`torch.Tensor`, *optional*):
+                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
+                provided, text embeddings will be generated from `prompt` input argument.
+            negative_prompt_embeds (`torch.Tensor`, *optional*):
+                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
+                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
+                argument.
+            pooled_prompt_embeds (`torch.Tensor`, *optional*):
+                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
+                If not provided, pooled text embeddings will be generated from `prompt` input argument.
+            negative_pooled_prompt_embeds (`torch.Tensor`, *optional*):
+                Pre-generated negative pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
+                weighting. If not provided, pooled negative_prompt_embeds will be generated from `negative_prompt`
+                input argument.
+            lora_scale (`float`, *optional*):
+                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
+            clip_skip (`int`, *optional*):
+                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
+                the output of the pre-final layer will be used for computing the prompt embeddings.
+        """
+        device = device or self._execution_device
+
+        # set lora scale so that monkey patched LoRA
+        # function of text encoder can correctly access it
+        if lora_scale is not None and isinstance(self, StableDiffusionXLLoraLoaderMixin):
+            self._lora_scale = lora_scale
+
+            # dynamically adjust the LoRA scale
+            if self.text_encoder is not None:
+                if not USE_PEFT_BACKEND:
+                    adjust_lora_scale_text_encoder(self.text_encoder, lora_scale)
+                else:
+                    scale_lora_layers(self.text_encoder, lora_scale)
+
+            if self.text_encoder_2 is not None:
+                if not USE_PEFT_BACKEND:
+                    adjust_lora_scale_text_encoder(self.text_encoder_2, lora_scale)
+                else:
+                    scale_lora_layers(self.text_encoder_2, lora_scale)
+
+        prompt = [prompt] if isinstance(prompt, str) else prompt
+
+        if prompt is not None:
+            batch_size = len(prompt)
+        else:
+            batch_size = prompt_embeds.shape[0]
+
+        # Define tokenizers and text encoders
+        tokenizers = [self.tokenizer, self.tokenizer_2] if self.tokenizer is not None else [self.tokenizer_2]
+        text_encoders = (
+            [self.text_encoder, self.text_encoder_2] if self.text_encoder is not None else [self.text_encoder_2]
+        )
+
+        if prompt_embeds is None:
+            prompt_2 = prompt_2 or prompt
+            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
+
+            # textual inversion: process multi-vector tokens if necessary
+            prompt_embeds_list = []
+            prompts = [prompt, prompt_2]
+            for prompt, tokenizer, text_encoder in zip(prompts, tokenizers, text_encoders):
+                if isinstance(self, TextualInversionLoaderMixin):
+                    prompt = self.maybe_convert_prompt(prompt, tokenizer)
+
+                text_inputs = tokenizer(
+                    prompt,
+                    padding="max_length",
+                    max_length=tokenizer.model_max_length,
+                    truncation=True,
+                    return_tensors="pt",
+                )
+
+                text_input_ids = text_inputs.input_ids
+                untruncated_ids = tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
+
+                if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
+                    text_input_ids, untruncated_ids
+                ):
+                    removed_text = tokenizer.batch_decode(untruncated_ids[:, tokenizer.model_max_length - 1 : -1])
+                    logger.warning(
+                        "The following part of your input was truncated because CLIP can only handle sequences up to"
+                        f" {tokenizer.model_max_length} tokens: {removed_text}"
+                    )
+
+                prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)
+
+                # We are only ALWAYS interested in the pooled output of the final text encoder
+                if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
+                    pooled_prompt_embeds = prompt_embeds[0]
+
+                if clip_skip is None:
+                    prompt_embeds = prompt_embeds.hidden_states[-2]
+                else:
+                    # "2" because SDXL always indexes from the penultimate layer.
+                    prompt_embeds = prompt_embeds.hidden_states[-(clip_skip + 2)]
+
+                prompt_embeds_list.append(prompt_embeds)
+
+            prompt_embeds = torch.concat(prompt_embeds_list, dim=-1)
+
+        # get unconditional embeddings for classifier free guidance
+        zero_out_negative_prompt = negative_prompt is None and self.config.force_zeros_for_empty_prompt
+        if do_classifier_free_guidance and negative_prompt_embeds is None and zero_out_negative_prompt:
+            negative_prompt_embeds = torch.zeros_like(prompt_embeds)
+            negative_pooled_prompt_embeds = torch.zeros_like(pooled_prompt_embeds)
+        elif do_classifier_free_guidance and negative_prompt_embeds is None:
+            negative_prompt = negative_prompt or ""
+            negative_prompt_2 = negative_prompt_2 or negative_prompt
+
+            # normalize str to list
+            negative_prompt = batch_size * [negative_prompt] if isinstance(negative_prompt, str) else negative_prompt
+            negative_prompt_2 = (
+                batch_size * [negative_prompt_2] if isinstance(negative_prompt_2, str) else negative_prompt_2
+            )
+
+            uncond_tokens: List[str]
+            if prompt is not None and type(prompt) is not type(negative_prompt):
+                raise TypeError(
+                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
+                    f" {type(prompt)}."
+                )
+            elif batch_size != len(negative_prompt):
+                raise ValueError(
+                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
+                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
+                    " the batch size of `prompt`."
+                )
+            else:
+                uncond_tokens = [negative_prompt, negative_prompt_2]
+
+            negative_prompt_embeds_list = []
+            for negative_prompt, tokenizer, text_encoder in zip(uncond_tokens, tokenizers, text_encoders):
+                if isinstance(self, TextualInversionLoaderMixin):
+                    negative_prompt = self.maybe_convert_prompt(negative_prompt, tokenizer)
+
+                max_length = prompt_embeds.shape[1]
+                uncond_input = tokenizer(
+                    negative_prompt,
+                    padding="max_length",
+                    max_length=max_length,
+                    truncation=True,
+                    return_tensors="pt",
+                )
+
+                negative_prompt_embeds = text_encoder(
+                    uncond_input.input_ids.to(device),
+                    output_hidden_states=True,
+                )
+
+                # We are only ALWAYS interested in the pooled output of the final text encoder
+                if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
+                    negative_pooled_prompt_embeds = negative_prompt_embeds[0]
+                negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]
+
+                negative_prompt_embeds_list.append(negative_prompt_embeds)
+
+            negative_prompt_embeds = torch.concat(negative_prompt_embeds_list, dim=-1)
+
+        if self.text_encoder_2 is not None:
+            prompt_embeds = prompt_embeds.to(dtype=self.text_encoder_2.dtype, device=device)
+        else:
+            prompt_embeds = prompt_embeds.to(dtype=self.unet.dtype, device=device)
+
+        bs_embed, seq_len, _ = prompt_embeds.shape
+        # duplicate text embeddings for each generation per prompt, using mps friendly method
+        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
+        prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)
+
+        if do_classifier_free_guidance:
+            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
+            seq_len = negative_prompt_embeds.shape[1]
+
+            if self.text_encoder_2 is not None:
+                negative_prompt_embeds = negative_prompt_embeds.to(dtype=self.text_encoder_2.dtype, device=device)
+            else:
+                negative_prompt_embeds = negative_prompt_embeds.to(dtype=self.unet.dtype, device=device)
+
+            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
+            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
+
+        pooled_prompt_embeds = pooled_prompt_embeds.repeat(1, num_images_per_prompt).view(
+            bs_embed * num_images_per_prompt, -1
+        )
+        if do_classifier_free_guidance:
+            negative_pooled_prompt_embeds = negative_pooled_prompt_embeds.repeat(1, num_images_per_prompt).view(
+                bs_embed * num_images_per_prompt, -1
+            )
+
+        if self.text_encoder is not None:
+            if isinstance(self, StableDiffusionXLLoraLoaderMixin) and USE_PEFT_BACKEND:
+                # Retrieve the original scale by scaling back the LoRA layers
+                unscale_lora_layers(self.text_encoder, lora_scale)
+
+        if self.text_encoder_2 is not None:
+            if isinstance(self, StableDiffusionXLLoraLoaderMixin) and USE_PEFT_BACKEND:
+                # Retrieve the original scale by scaling back the LoRA layers
+                unscale_lora_layers(self.text_encoder_2, lora_scale)
+
+        return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_image
+    def encode_image(self, image, device, num_images_per_prompt, output_hidden_states=None):
+        dtype = next(self.image_encoder.parameters()).dtype
+
+        if not isinstance(image, torch.Tensor):
+            image = self.feature_extractor(image, return_tensors="pt").pixel_values
+
+        image = image.to(device=device, dtype=dtype)
+        if output_hidden_states:
+            image_enc_hidden_states = self.image_encoder(image, output_hidden_states=True).hidden_states[-2]
+            image_enc_hidden_states = image_enc_hidden_states.repeat_interleave(num_images_per_prompt, dim=0)
+            uncond_image_enc_hidden_states = self.image_encoder(
+                torch.zeros_like(image), output_hidden_states=True
+            ).hidden_states[-2]
+            uncond_image_enc_hidden_states = uncond_image_enc_hidden_states.repeat_interleave(
+                num_images_per_prompt, dim=0
+            )
+            return image_enc_hidden_states, uncond_image_enc_hidden_states
+        else:
+            image_embeds = self.image_encoder(image).image_embeds
+            image_embeds = image_embeds.repeat_interleave(num_images_per_prompt, dim=0)
+            uncond_image_embeds = torch.zeros_like(image_embeds)
+
+            return image_embeds, uncond_image_embeds
+
+    def upcast_vae(self):
+        dtype = self.vae.dtype
+        self.vae.to(dtype=torch.float32)
+        use_torch_2_0_or_xformers = isinstance(
+            self.vae.decoder.mid_block.attentions[0].processor,
+            (AttnProcessor2_0, XFormersAttnProcessor, FusedAttnProcessor2_0),
+        )
+        # if xformers or torch_2_0 is used attention block does not need
+        # to be in float32 which can save lots of memory
+        if use_torch_2_0_or_xformers:
+            self.vae.post_quant_conv.to(dtype)
+            self.vae.decoder.conv_in.to(dtype)
+            self.vae.decoder.mid_block.to(dtype)
+
+    # Copied from diffusers.pipelines.latent_consistency_models.pipeline_latent_consistency_text2img.LatentConsistencyModelPipeline.get_guidance_scale_embedding
+    def get_guidance_scale_embedding(
+        self, w: torch.Tensor, embedding_dim: int = 512, dtype: torch.dtype = torch.float32
+    ) -> torch.Tensor:
+        """
+        See https://github.com/google-research/vdm/blob/dc27b98a554f65cdc654b800da5aa1846545d41b/model_vdm.py#L298
+
+        Args:
+            w (`torch.Tensor`):
+                Generate embedding vectors with a specified guidance scale to subsequently enrich timestep embeddings.
+            embedding_dim (`int`, *optional*, defaults to 512):
+                Dimension of the embeddings to generate.
+            dtype (`torch.dtype`, *optional*, defaults to `torch.float32`):
+                Data type of the generated embeddings.
+
+        Returns:
+            `torch.Tensor`: Embedding vectors with shape `(len(w), embedding_dim)`.
+        """
+        assert len(w.shape) == 1
+        w = w * 1000.0
+
+        half_dim = embedding_dim // 2
+        emb = torch.log(torch.tensor(10000.0)) / (half_dim - 1)
+        emb = torch.exp(torch.arange(half_dim, dtype=dtype) * -emb)
+        emb = w.to(dtype)[:, None] * emb[None, :]
+        emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=1)
+        if embedding_dim % 2 == 1:  # zero pad
+            emb = torch.nn.functional.pad(emb, (0, 1))
+        assert emb.shape == (w.shape[0], embedding_dim)
+        return emb
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
+    def prepare_extra_step_kwargs(self, generator, eta):
+        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
+        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
+        # eta corresponds to η in DDIM paper: https://huggingface.co/papers/2010.02502
+        # and should be between [0, 1]
+
+        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
+        extra_step_kwargs = {}
+        if accepts_eta:
+            extra_step_kwargs["eta"] = eta
+
+        # check if the scheduler accepts generator
+        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
+        if accepts_generator:
+            extra_step_kwargs["generator"] = generator
+        return extra_step_kwargs
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_ip_adapter_image_embeds
+    def prepare_ip_adapter_image_embeds(
+        self, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt, do_classifier_free_guidance
+    ):
+        image_embeds = []
+        if do_classifier_free_guidance:
+            negative_image_embeds = []
+        if ip_adapter_image_embeds is None:
+            if not isinstance(ip_adapter_image, list):
+                ip_adapter_image = [ip_adapter_image]
+
+            if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
+                raise ValueError(
+                    f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
+                )
+
+            for single_ip_adapter_image, image_proj_layer in zip(
+                ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
+            ):
+                output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
+                single_image_embeds, single_negative_image_embeds = self.encode_image(
+                    single_ip_adapter_image, device, 1, output_hidden_state
+                )
+
+                image_embeds.append(single_image_embeds[None, :])
+                if do_classifier_free_guidance:
+                    negative_image_embeds.append(single_negative_image_embeds[None, :])
+        else:
+            for single_image_embeds in ip_adapter_image_embeds:
+                if do_classifier_free_guidance:
+                    single_negative_image_embeds, single_image_embeds = single_image_embeds.chunk(2)
+                    negative_image_embeds.append(single_negative_image_embeds)
+                image_embeds.append(single_image_embeds)
+
+        ip_adapter_image_embeds = []
+        for i, single_image_embeds in enumerate(image_embeds):
+            single_image_embeds = torch.cat([single_image_embeds] * num_images_per_prompt, dim=0)
+            if do_classifier_free_guidance:
+                single_negative_image_embeds = torch.cat([negative_image_embeds[i]] * num_images_per_prompt, dim=0)
+                single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds], dim=0)
+
+            single_image_embeds = single_image_embeds.to(device=device)
+            ip_adapter_image_embeds.append(single_image_embeds)
+
+        return ip_adapter_image_embeds
+
+    @property
+    def guidance_scale(self):
+        return self._guidance_scale
+
+    @property
+    def guidance_rescale(self):
+        return self._guidance_rescale
+
+    @property
+    def clip_skip(self):
+        return self._clip_skip
+
+    # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
+    # of the Imagen paper: https://huggingface.co/papers/2205.11487 . `guidance_scale = 1`
+    # corresponds to doing no classifier free guidance.
+    @property
+    def do_classifier_free_guidance(self):
+        return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
+
+    @property
+    def cross_attention_kwargs(self):
+        return self._cross_attention_kwargs
+
+    @property
+    def denoising_end(self):
+        return self._denoising_end
+
+    @property
+    def denoising_start(self):
+        return self._denoising_start
+
+    @property
+    def num_timesteps(self):
+        return self._num_timesteps
+
+    @property
+    def interrupt(self):
+        return self._interrupt
diff --git a/src/diffusers/pipelines/stable_video_diffusion/pipeline_stable_video_diffusion.py b/src/diffusers/pipelines/stable_video_diffusion/pipeline_stable_video_diffusion.py
index 6d9053faaec8..3090202414cb 100644
--- a/src/diffusers/pipelines/stable_video_diffusion/pipeline_stable_video_diffusion.py
+++ b/src/diffusers/pipelines/stable_video_diffusion/pipeline_stable_video_diffusion.py
@@ -70,7 +70,7 @@ def _append_dims(x, target_dims):
     return x[(...,) + (None,) * dims_to_append]
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/src/diffusers/pipelines/t2i_adapter/pipeline_stable_diffusion_adapter.py b/src/diffusers/pipelines/t2i_adapter/pipeline_stable_diffusion_adapter.py
index 1ce6987114a7..455c133d8b4d 100644
--- a/src/diffusers/pipelines/t2i_adapter/pipeline_stable_diffusion_adapter.py
+++ b/src/diffusers/pipelines/t2i_adapter/pipeline_stable_diffusion_adapter.py
@@ -128,7 +128,7 @@ def _preprocess_adapter_image(image, height, width):
     return image
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -274,7 +274,7 @@ def __init__(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -307,7 +307,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -490,7 +490,7 @@ def encode_prompt(
 
         return prompt_embeds, negative_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -505,7 +505,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -517,7 +517,7 @@ def decode_latents(self, latents):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/t2i_adapter/pipeline_stable_diffusion_xl_adapter.py b/src/diffusers/pipelines/t2i_adapter/pipeline_stable_diffusion_xl_adapter.py
index 2802d690f3cc..4b5c2d2b1b79 100644
--- a/src/diffusers/pipelines/t2i_adapter/pipeline_stable_diffusion_xl_adapter.py
+++ b/src/diffusers/pipelines/t2i_adapter/pipeline_stable_diffusion_xl_adapter.py
@@ -126,7 +126,7 @@ def _preprocess_adapter_image(image, height, width):
     return image
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -153,7 +153,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -310,7 +310,7 @@ def __init__(
             else 128
         )
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -620,7 +620,7 @@ def prepare_ip_adapter_image_embeds(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -783,10 +783,7 @@ def upcast_vae(self):
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
diff --git a/src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_synth.py b/src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_synth.py
index 3ce7b4d1990f..64fb59d0e30a 100644
--- a/src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_synth.py
+++ b/src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_synth.py
@@ -122,7 +122,7 @@ def __init__(
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.video_processor = VideoProcessor(do_resize=False, vae_scale_factor=self.vae_scale_factor)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -155,7 +155,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -350,7 +350,7 @@ def decode_latents(self, latents):
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_synth_img2img.py b/src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_synth_img2img.py
index 9d0b7e3dbc32..efb8a5c85b99 100644
--- a/src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_synth_img2img.py
+++ b/src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_synth_img2img.py
@@ -88,7 +88,7 @@
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -157,7 +157,7 @@ def __init__(
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.video_processor = VideoProcessor(do_resize=False, vae_scale_factor=self.vae_scale_factor)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -190,7 +190,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -386,7 +386,7 @@ def decode_latents(self, latents):
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
diff --git a/src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_zero.py b/src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_zero.py
index 96316f8e91e5..21b8bf6676ef 100644
--- a/src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_zero.py
+++ b/src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_zero.py
@@ -780,7 +780,7 @@ def __call__(
 
         return TextToVideoPipelineOutput(images=image, nsfw_content_detected=has_nsfw_concept)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -795,7 +795,7 @@ def run_safety_checker(self, image, device, dtype):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -813,7 +813,7 @@ def prepare_extra_step_kwargs(self, generator, eta):
             extra_step_kwargs["generator"] = generator
         return extra_step_kwargs
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
diff --git a/src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_zero_sdxl.py b/src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_zero_sdxl.py
index 288aae6c0d44..ac365e638d21 100644
--- a/src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_zero_sdxl.py
+++ b/src/diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_zero_sdxl.py
@@ -318,7 +318,7 @@ def create_motion_field_and_warp_latents(motion_field_strength_x, motion_field_s
     return warped_latents
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -437,7 +437,7 @@ def __init__(
         else:
             self.watermark = None
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -455,17 +455,13 @@ def prepare_extra_step_kwargs(self, generator, eta):
             extra_step_kwargs["generator"] = generator
         return extra_step_kwargs
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.upcast_vae
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.upcast_vae
     def upcast_vae(self):
         dtype = self.vae.dtype
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-                FusedAttnProcessor2_0,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor, FusedAttnProcessor2_0),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
@@ -595,7 +591,7 @@ def check_inputs(
                 "If `negative_prompt_embeds` are provided, `negative_pooled_prompt_embeds` also have to be passed. Make sure to generate `negative_pooled_prompt_embeds` from the same text encoder that was used to generate `negative_prompt_embeds`."
             )
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
diff --git a/src/diffusers/pipelines/unidiffuser/pipeline_unidiffuser.py b/src/diffusers/pipelines/unidiffuser/pipeline_unidiffuser.py
index f9298d5b86f8..59d927506203 100644
--- a/src/diffusers/pipelines/unidiffuser/pipeline_unidiffuser.py
+++ b/src/diffusers/pipelines/unidiffuser/pipeline_unidiffuser.py
@@ -150,7 +150,7 @@ def __init__(
         # TODO: handle safety checking?
         self.safety_checker = None
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -384,7 +384,7 @@ def _infer_batch_size(
                 )
         return batch_size, multiplier
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -417,7 +417,7 @@ def _encode_prompt(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt with self.tokenizer->self.clip_tokenizer
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt with self.tokenizer->self.clip_tokenizer
     def encode_prompt(
         self,
         prompt,
diff --git a/src/diffusers/pipelines/visualcloze/pipeline_visualcloze_generation.py b/src/diffusers/pipelines/visualcloze/pipeline_visualcloze_generation.py
index e12995106bcf..5039f7ddacec 100644
--- a/src/diffusers/pipelines/visualcloze/pipeline_visualcloze_generation.py
+++ b/src/diffusers/pipelines/visualcloze/pipeline_visualcloze_generation.py
@@ -189,7 +189,7 @@ def __init__(
         )
         self.default_sample_size = 128
 
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_t5_prompt_embeds
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._get_t5_prompt_embeds
     def _get_t5_prompt_embeds(
         self,
         prompt: Union[str, List[str]] = None,
@@ -239,7 +239,7 @@ def _get_t5_prompt_embeds(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._get_clip_prompt_embeds
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._get_clip_prompt_embeds
     def _get_clip_prompt_embeds(
         self,
         prompt: Union[str, List[str]],
@@ -284,7 +284,7 @@ def _get_clip_prompt_embeds(
 
         return prompt_embeds
 
-    # Modified from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_prompt
+    # Modified from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin.encode_prompt
     def encode_prompt(
         self,
         layout_prompt: Union[str, List[str]],
@@ -488,7 +488,7 @@ def _prepare_latent_image_ids(image, vae_scale_factor, device, dtype):
         return torch.cat(latent_image_ids, dim=0)
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._pack_latents
     def _pack_latents(latents, batch_size, num_channels_latents, height, width):
         latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
         latents = latents.permute(0, 2, 4, 1, 3, 5)
diff --git a/src/diffusers/pipelines/wan/pipeline_wan_animate.py b/src/diffusers/pipelines/wan/pipeline_wan_animate.py
index c7c983b2f7d4..e053948c4329 100644
--- a/src/diffusers/pipelines/wan/pipeline_wan_animate.py
+++ b/src/diffusers/pipelines/wan/pipeline_wan_animate.py
@@ -133,7 +133,7 @@ def prompt_clean(text):
     return text
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/pipelines/wan/pipeline_wan_i2v.py b/src/diffusers/pipelines/wan/pipeline_wan_i2v.py
index b7fd0b05980f..fa1bd9092c3b 100644
--- a/src/diffusers/pipelines/wan/pipeline_wan_i2v.py
+++ b/src/diffusers/pipelines/wan/pipeline_wan_i2v.py
@@ -110,7 +110,7 @@ def prompt_clean(text):
     return text
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/pipelines/wan/pipeline_wan_vace.py b/src/diffusers/pipelines/wan/pipeline_wan_vace.py
index 351ae2e70563..8ae92a6f0fcc 100644
--- a/src/diffusers/pipelines/wan/pipeline_wan_vace.py
+++ b/src/diffusers/pipelines/wan/pipeline_wan_vace.py
@@ -124,7 +124,7 @@ def prompt_clean(text):
     return text
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/pipelines/wan/pipeline_wan_video2video.py b/src/diffusers/pipelines/wan/pipeline_wan_video2video.py
index a976126da7fe..53b3aada4b35 100644
--- a/src/diffusers/pipelines/wan/pipeline_wan_video2video.py
+++ b/src/diffusers/pipelines/wan/pipeline_wan_video2video.py
@@ -97,7 +97,7 @@ def prompt_clean(text):
     return text
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -157,7 +157,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
diff --git a/src/diffusers/pipelines/z_image/pipeline_z_image.py b/src/diffusers/pipelines/z_image/pipeline_z_image.py
index a4fcacb6eb9b..e0eb0ea67d1c 100644
--- a/src/diffusers/pipelines/z_image/pipeline_z_image.py
+++ b/src/diffusers/pipelines/z_image/pipeline_z_image.py
@@ -60,7 +60,7 @@
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -74,7 +74,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
diff --git a/tests/pipelines/stable_diffusion/test_stable_diffusion.py b/tests/pipelines/stable_diffusion/test_stable_diffusion.py
index c9d9525b2e45..d463c77b0a2e 100644
--- a/tests/pipelines/stable_diffusion/test_stable_diffusion.py
+++ b/tests/pipelines/stable_diffusion/test_stable_diffusion.py
@@ -538,7 +538,7 @@ def test_stable_diffusion_long_prompt(self):
         do_classifier_free_guidance = True
         negative_prompt = None
         num_images_per_prompt = 1
-        logger = logging.get_logger("diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion")
+        logger = logging.get_logger("diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils")
         logger.setLevel(logging.WARNING)
 
         prompt = 100 * "@"
diff --git a/tests/pipelines/stable_diffusion_2/test_stable_diffusion.py b/tests/pipelines/stable_diffusion_2/test_stable_diffusion.py
index 3b2552b432d3..3d58bb2774ac 100644
--- a/tests/pipelines/stable_diffusion_2/test_stable_diffusion.py
+++ b/tests/pipelines/stable_diffusion_2/test_stable_diffusion.py
@@ -272,7 +272,7 @@ def test_stable_diffusion_long_prompt(self):
         do_classifier_free_guidance = True
         negative_prompt = None
         num_images_per_prompt = 1
-        logger = logging.get_logger("diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion")
+        logger = logging.get_logger("diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils")
         logger.setLevel(logging.WARNING)
 
         prompt = 25 * "@"