Skip to content

wangqiqi12/MagicOmini

Repository files navigation

MagicOmini: A Magic Demo for You to Enjoy Local Sketch Editing


MagicOmini: Efficient Local Sketch Editing for Diffusion Transformers
Qiqi Wang, University of Chinese Academy of Sciences

Features

MagicOmini is a minimal yet powerful local sketch-editing framework for Diffusion Transformer models like FLUX, built upon OminiControl.

Quick Start

Setup (Optional)

  1. Environment setup
conda create -n magicomini python=3.11
conda activate magicomini
  1. Requirements installation
pip install torch==2.4.1 torchvision==0.19.1 torchaudio==2.4.1 --index-url https://download.pytorch.org/whl/cu124
pip install -r requirements.txt
  1. Checkpoints

Click here and download the checkpoint. Remember to change the checkpoint to your path in gradio code.

Visual Result

Simple UI sketch example

Complicated UI sketch example

UI sketch example with hint of color

Different color

Same color of different intensity

Notation

  1. Input images are automatically resized to 512x512 or 1024x1024 resolution.

Acknowledgment

We would like to thank the OminiControl team for their excellent work, which serves as the foundation for this project.

Citation

If you find this work useful, please consider citing:

@article{ominicontrol,
  title={OminiControl: Minimal and Universal Control for Diffusion Transformer},
  author={Xiang, Zhenxiong and Hu, Xinzhe and Zheng, Haoji and Dang, Yingqing and Li, Ying and Shi, Min and Liu, Haoming and Qiao, Yu and Shan, Ying},
  journal={arXiv preprint arXiv:2411.15098},
  year={2024}
}

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors