Home

Awesome

InteractDiffusion for Automatic1111 Stable Diffusion WebUI

Warning: experimental inital implementation, feel free to provide feedback and pull request to improve this extension.

This extension is designed for AUTOMATIC1111's Stable Diffusion web UI, enabling the integration of InteractDiffusion with the original Stable Diffusion model. This integration enhances the capability of the web UI to control the interactions in image generation. It works with DreamBooth and LoRA models.

https://github.com/jiuntian/sd-webui-interactdiffusion/assets/13869695/7ea6d9f5-197a-46b2-b3c5-df659aa97f07

Installation

  1. Install the extension via link.

    Open "Extensions" tab and then select "Install from URL". Enter https://github.com/jiuntian/sd-webui-interactdiffusion.git to "URL for extension's git repository" and click Install. Finally, click "Apply and restart UI".

  2. Download the model at HuggingFace Hub with name ext_interactdiff_v1.2.pth.

  3. Put models in "stable-diffusion-webui\extensions\sd-webui-interactdiffusion\models".

How to Use

  1. Enable the extension by checking the Enabled.
  2. Please append comma seperator to end of each entry in grounding instructions like "tohru=feeding=cat;a=doing=b".
  3. Click on Create Drawing Canvas to create a new canvas.
  4. Draw correspinding bounding boxes for subject and object, and verify the interactions on the right side.
  5. Adjust scheduling sampling when necessarily.
  6. Generate the image as usual.

Gallery

Some examples generated with InteractDiffusion, together with other DreamBooth and LoRA models.

    
image (7)image (5)image (6)image (4)
cuteyukimix_1cuteyukimix_7darksushimix_1toonyou_6
image (8)cuteyukimix_4darksushimix_5rcnzcartoon_1

Limitations

  1. We currently do not support SDXL yet due to limited computation resources.
  2. It could possibly cause conflicts with other extensions.
  3. Some artefacts could happens on some LoRA models.

Citation

@misc{hoe2023interactdiffusion,
      title={InteractDiffusion: Interaction Control in Text-to-Image Diffusion Models}, 
      author={Jiun Tian Hoe and Xudong Jiang and Chee Seng Chan and Yap-Peng Tan and Weipeng Hu},
      year={2023},
      eprint={2312.05849},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

Related Projects

This implementation is constructed based on the foundation of sd_webui_gligen.