Awesome

FlowEdit

Project | Arxiv | Demo | ComfyUI

Official Pytorch implementation of the paper: "FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models"

Installation

1. Clone the repository

2. Install the required dependencies using pip install torch diffusers transformers accelerate sentencepiece protobuf <br>

   • Tested with CUDA version 12.4 and diffusers 0.31.0

Running examples

Run editing with Stable Diffusion 3: python run_script.py --exp_yaml SD3_exp.yaml

Run editing with Flux: python run_script.py --exp_yaml FLUX_exp.yaml

Usage - your own examples

• Upload images to example_images folder.

• Create an edits file that specifies: (a) a path to the input image, (b) a source prompt, (c) target prompts, and (d) target codes. The target codes summarize the changes between the source and target prompts and will appear in the output filename. <br> See edits.yaml for example.

• Create an experiment file containing the hyperparamaters needed for running FlowEdit, such as n_max, n_min. This file also includes the path to the edits.yaml file<br> See FLUX_exp.yaml for FLUX usage example and SD3_exp.yaml for Stable Diffusion 3 usage example. <br> For a detailed discussion on the impact of different hyperparameters and the values we used, please refer to our paper.

Run python run_script.py --exp_yaml <path to your experiment yaml>

ComfyUI implementation for different models

• FLUX
• LTX Video
• HunyuanLoom

Implemented by logtd

License

This project is licensed under the MIT License.

Citation

If you use this code for your research, please cite our paper:

@article{kulikov2024flowedit,
	title = {FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models},
	author = {Kulikov, Vladimir and Kleiner, Matan and Huberman-Spiegelglas, Inbar and Michaeli, Tomer},
	journal = {arXiv preprint arXiv:2412.08629},
	year = {2024}
	}