Could use guidance on how to train about riffusion HOT 3 CLOSED

Hey @dustyny! Yes you can use any standard training or dreambooth method as long as you have compatible spectrogram images with text. And you can use the CLI example to help convert with your desired parameters. Whether you fine tune or use dreambooth depends on how much data you have, typically dreambooth is for a much smaller dataset than hundreds of thousands of files.

The model can be imported into riffusion, and you can use it within automatic1111 (an extension to convert to audio also exists). Img2img, outpainting, and so on all work.

from riffusion.

I'd be keen to hear about your use case if you DM me on discord as well.

from riffusion.

I really appreciate the response, I'll connect with you on discord to discuss more in depth. It sounds like standard training (I want to understand what you mean by that) is my best approach, I have a lot of data that I can use for training.

from riffusion.