Thank you! :)

[asigalov61]

@yoyolicoris Hey bro!

First of all, I wanted to thank you for this repo. It was very educational to me and I the pre-trained models are fantastic :)

The reason why I am reaching out to you is because I wanted to ask you for advice and possible some help with my fork of your repo.

If you have a minute, please take a look:

https://github.com/asigalov61/neuralpiano

I basically want to create a solo piano model for MIDI synthesis (think neural SF) but I am having problems getting good results.

I was able to refactor and repurpose the code. It works and trains fine as far as I can tell, however, the inference does not produce good results even after many epochs.

I am attaching a sample plot, audio output and a log file for you to check out as well.

logs.zip

I am new to spectrogram diffusion models/implementations so I could really use your advice and help here :)

Please reply at your convenience.

Sincerely,

Alex.

[yoyolicoris]

Hi, happy to hear that you find my repo helpful to you :)
I'll have a look when I'm available.

[asigalov61]

@yoyolicoris Thank you. Let me know, please!