Riffusion is an app for real-time music generation with stable diffusion. It is based on audio diffusion↗︎. With diffusion models, it is possible to condition their creations not only on a text prompt but also on other images. It uses the v1.5 stable diffusion model with no modifications, just fine-tuned on images of spectrograms paired with text. It can generate infinite variations of a prompt by varying the seed. All the same web UIs and techniques like img2img, inpainting, negative prompts, and interpolation work out of the box.