README.md

---
license: mit
task_categories:
- image-to-video
language:
- en
tags:
- diffusion
- generation
- human
- animation
size_categories:
- 10K<n<100K
---

## Data for Human Image Animation

<img src="https://github.com/user-attachments/assets/01174ec4-c076-4947-966c-01d511d0383e">

## 🎏 Introduction
<b>TL; DR: With the rapid developments in generative models, including the diffusion-based or the flow-based models, the human-centric tasks, like pose-driven human image animation, audio-driven action generation,  diffusion-based pose estimation, human optical estimation, etc., have attracted a lot of attention from lots of works.

We pay attention to the quality of the training data of human data for these tasks. However, due to the lack of high-quality datasets, especially for the human image animation, we find it is hard to collect videos from existing public datasets, while these videos have these characteristics:

1. High-resolution: the resolution of the vertical video is larger than 1080 * 576.
2. High-dynamic: the video is vivid and suitable to learn human motions.
3. Dancing-style: In this stage, we focus on the human animation task and mainly collect videos like TikTok styles.

## ⚔️ What we do

We collect a large number of videos from the internet. After filtering low-quality, limited motion, and bad frames, we get 25,000 videos in this repo. Now we provide a visualization to these data and the corresponding pose data, you can check each training video in our work.

Notice: we do not allow any commercial usage of these videos and you must delete them within 24 hours after downloading.

Tips: If you find that your data is being infringed upon, please contact us immediately to request its removal.

## 📍 Citation 

If you find this guidance helpful, please consider citing:

```
@article{zhao2025dynamictrl,
      title={DynamiCtrl: Rethinking the Basic Structure and the Role of Text for High-quality Human Image Animation}, 
      author={Haoyu, Zhao and Zhongang, Qi and Cong, Wang and Qingping, Zheng and Guansong, Lu and Fei, Chen and Hang, Xu and Zuxuan, Wu},
      year={2025},
      journal={arXiv:2503.21246},
}
```