This inexplicably blew up on HN a few weeks ago.

I personally found the generated audio not all that great especially classical. Clever idea though

Riffusion – Stable Diffusion fine-tuned to generate music https://news.ycombinator.com/item?id=33999162

Riffusion is an excellent way to be flexible with music production. An artist can go from heavy metal to pop. Not to mention, using stable diffusion and interpolation in latent space will create smooth transitions in the generated audio clips, which could help to create more coherent and pleasant-sounding music. I'm curious about one thing, though. What do you use for semantic analysis of the text inputs? Have you had issues with that?

Hey with the rate with how bad and repetitive music is getting it might replace it in some instances.

So true !

I'm impressed that it worked at all, with how different spectrograms are from natural images.

I made a review:

https://www.youtube.com/watch?v=_UE0T7ll4rA