Discord bot and Interface for Stable Diffusion

Github Project / Download

Project information and download links are available here

Below are examples of out-painting images made using the G-Diffuser Discord bot; the original image is on the left and the expanded / out-painted image is on the right.

The complete pipeline / stack used is as follows:
  • RunwayML Stable Diffusion 1.5 w/in-painting u-net and upgraded VAE
  • Latent space shaped noise
  • CLIP guidance

CLIP interrogation is the only part of the process that is currently not automatic / integrated, but it is my intention to integrate it into the bot directly in the near future.

Join us on the Stable Cabal Discord