Thank you!

#3
by asigalov61 - opened

@drscotthawley

Hey Scott,

Thanks a lot of making this great demo :) Its very nice and I enjoyed it very much :)

I wanted to suggest that you look into img2img diffusion pipeline/method. This can create very nice results too.

Here is the HF space you can try with the demo image I have attached:
https://huggingface.co/spaces/diffusers/unofficial-SDXL-Turbo-i2i-t2i

download.png

I think it maybe interesting to try in terms of generating remixes of the song (chunks). An inpainting of sorts.
And I think you can easily adopt the code from SDXL img2img pipeline to your model so that it generates more suitable and more correct MIDI matrixes/images.

Otherwise, thanks again for this demo/project.

If you like, I would be happy to chat more.

Let me know.

Alex.

Sign up or log in to comment