EDiffi, Nvidia's new State-of-the-Art Diffusion model

Mixture of expert denoisers (basically multiple diffusion models for different noise levels rather than a single model for the entire diffusion process) might become the next big thing in diffusion model literature: https://twitter.com/iScienceLuvr/status/1587973173932195840

Allows you to have "fill in the blank" img2img ! ! !

- Use basic colors to fill out desired "layout" of image

- Label the image (type or scrible) with "dog", "sunset", "Dungeons and Dragons", etc

- Those parts of photo now bised towards those subjects

Highlights 1.Beats Imagen in quality 2.Incredible img+text2img capabilities