v0xie/sd-webui-cads

Greatly increase the diversity of your generated images in Automatic1111 WebUI through Condition-Annealed Sampling.

sd-webui-cads

An implementation of the method in CADS: Unleashing the Diversity of Diffusion Models through Condition-Annealed Sampling in Automatic1111 WebUI.

CADS greatly increases diversity of generated images by adding scheduled noise to the conditioning at inference time.

PR's are welcome!

Feature / To-do List

• SD XL support
• SD 1.5 support
• Hi-res fix support
• Support restoring parameter values from infotext (Send to Txt2Img, Send to Img2Img, etc.)
• Write infotext to image grids
• X/Y/Z plot support
• ControlNet support

Credits

• The authors of the original paper for their method (https://arxiv.org/abs/2310.17347):

  @inproceedings{
      sadat2024cads,
      title={{CADS}: Unleashing the Diversity of Diffusion Models through Condition-Annealed Sampling},
      author={Seyedmorteza Sadat and Jakob Buhmann and Derek Bradley and Otmar Hilliges and Romann M. Weber},
      booktitle={The Twelfth International Conference on Learning Representations},
      year={2024},
      url={https://openreview.net/forum?id=zMoNrajk2X}
  }
  

• @udon-universe's extension templates (https://github.com/udon-universe/stable-diffusion-webui-extension-templates)