A diffusion-based model for upscaling images to higher resolution, trained by Katherine Crowson in collaboration with Stability AI.

A technique for doing text-to-image synthesis cheaply using pretrained CLIP and diffusion models.

A highly efficient text-to-image model that can be trained without captioned images.

A trainable, memory-efficient, and GPU-friendly PyTorch reproduction of AlphaFold2

A dataset of prompts, synthetic AI generated images, and aesthetic ratings of those images.

A technique for doing text-to-image synthesis cheaply using pretrained CLIP and VQGAN models.