[NIPS 2024] Official code for FastDrag

FastDrag: Manipulate Anything in One Step

Xuanjia Zhao    Jian Guan    Congyi Fan    Dongli Xu   
Youtian Lin    Haiwei Pan    Pengming Feng

Installation

To install the required libraries, simply run the following command:

conda env create -f environment.yaml
conda activate fastdrag

Config

If you want download huggingface weights in local, you should download runwayml/stable-diffusion-v1-5 and SimianLuo/LCM_Dreamshaper_v7.

• Suggestion 1: It is suggested that download the model into the directory "local_pretrained_models";
• Suggestion 2: runwayml/stable-diffusion-v1-5 might not exist in huggingface, but can be found in other websites like gitee.

Then you can set path in config as below:

Run Fastdrag

To start with, in command line, run the following to start the gradio user interface:

python drag_ui.py

For users struggling in loading models from huggingface due to internet constraint, please run:

sh run_drag.sh

License

Code related to the FastDrag algorithm is under Apache 2.0 license.

BibTeX

If you find our repo helpful, please consider leaving a star or cite our paper :)

@inproceedings{
        zhao2024fastdrag,
        title={FastDrag: Manipulate Anything in One Step},
        author={Xuanjia Zhao and Jian Guan and Congyi Fan and Dongli Xu and Youtian Lin and Haiwei Pan and Pengming Feng},
        booktitle={The Thirty-eighth Annual Conference on Neural Information Processing Systems},
        year={2024}
}

Acknowledgement

The code is built based on DragDiffusion and diffusers, thanks for their outstanding work!

Notice

For the fisrt time to run, it may be slow, but it will perform normally afterwards.