Video-Motion-Customization (CVPR 2024)

This repository is the official implementation of VMC.
[CVPR 2024] VMC: Video Motion Customization using Temporal Attention Adaption for Text-to-Video Diffusion Models
Hyeonho Jeong*, Geon Yeong Park*, Jong Chul Ye,

Given an input video with any type of motion patterns, our framework, VMC fine-tunes only the Keyframe Generation Module within hierarchical Video Diffusion Models for motion-customized video generation.

News

• [2023.11.30] Initial Code Release (Additional codes will be uploaded.)

Setup

Requirements

pip install -r requirements.txt

Usage

The following command will run "train & inference" at the same time:

accelerate launch train_inference.py --config configs/car_forest.yml

*Additional scripts of 'train_only' and 'inference_with_pretrained' will be uploaded too.

Data

• PNG files: Google Drive Folder
• GIF files: Google Drive Folder

Results

Video Style Transfer

Backward Motion Customization

Citation

If you find our work interesting, please cite our paper.

@article{jeong2023vmc,
  title={VMC: Video Motion Customization using Temporal Attention Adaption for Text-to-Video Diffusion Models},
  author={Jeong, Hyeonho and Park, Geon Yeong and Ye, Jong Chul},
  journal={arXiv preprint arXiv:2312.00845},
  year={2023}
}

Shoutouts

• VMC directly employs an open-source project on cascaded Video Diffusion Models, Show-1, along with DeepFloyd IF.
• This code builds upon Diffusers and we referenced the code logic of Tune-A-Video.
• We conducted evaluation against 4 great projects: VideoComposer, Gen-1, Tune-A-Video, Control-A-Video

Thanks all for open-sourcing!