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.
Installation
-
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".
-
Download the model at HuggingFace Hub with name
ext_interactdiff_v1.2.pth
. -
Put models in "stable-diffusion-webui\extensions\sd-webui-interactdiffusion\models".
How to Use
- Enable the extension by checking the
Enabled
. - Please append comma seperator to end of each entry in grounding instructions like "tohru=feeding=cat;a=doing=b".
- Click on
Create Drawing Canvas
to create a new canvas. - Draw correspinding bounding boxes for subject and object, and verify the interactions on the right side.
- Adjust scheduling sampling when necessarily.
- Generate the image as usual.
Gallery
Some examples generated with InteractDiffusion, together with other DreamBooth and LoRA models.
Limitations
- We currently do not support SDXL yet due to limited computation resources.
- It could possibly cause conflicts with other extensions.
- 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.