VAP-Data

<div align="center">

# Video-As-Prompt: Unified Semantic Control for Video Generation

</div>

<div align="center">
<a href=https://bytedance.github.io/Video-As-Prompt target="_blank"><img src=https://img.shields.io/badge/Project%20Page-333399.svg?logo=homepage height=22px></a>
<a href=https://huggingface.co/collections/ByteDance/video-as-prompt target="_blank"><img src=https://img.shields.io/badge/%F0%9F%A4%97%20Models-d96902.svg height=22px></a>
<a href=https://huggingface.co/datasets/BianYx/VAP-Data target="_blank"><img src=https://img.shields.io/badge/%F0%9F%A4%97%20Dataset-276cb4.svg height=22px></a>
<a href=https://github.com/bytedance/Video-As-Prompt target="_blank"><img src= https://img.shields.io/badge/Code-black.svg?logo=github height=22px></a>
<a href=https://arxiv.org/pdf/2510.20888 target="_blank"><img src=https://img.shields.io/badge/Arxiv-b5212f.svg?logo=arxiv height=22px></a>
<!-- <a href=https://yxbian23.github.io/ target="_blank"><img src=https://img.shields.io/badge/Twitter-grey.svg?logo=x height=22px></a> -->
<!-- <a href="https://opensource.org/licenses/Apache">
<img src="https://img.shields.io/badge/License-Apache%202.0-lightgray">
</a> -->
<a href="https://www.youtube.com/watch?v=S3zpLIMOU4c" target="_blank">
<img src="https://img.shields.io/badge/%E2%96%B6%20YouTube%20Demo-FF0000.svg?logo=youtube&logoColor=white" height="24px">
</a>
</div>

<br>

## 🔥 News

- Oct 24, 2025: 📖 We release the first unified semantic video generation model, [Video-As-Prompt (VAP)](https://github.com/bytedance/Video-As-Prompt)!
- Oct 24, 2025: 🤗 We release the [VAP-Data](https://huggingface.co/datasets/BianYx/VAP-Data), the largest semantic-controlled video generation datasets with more than $100K$ samples!
- Oct 24, 2025: 👋 We present the [technical report](https://arxiv.org/pdf/2510.20888) of Video-As-Prompt, please check out the details and spark some discussion!

## 🖌️ **Video-As-Prompt**

> **Core idea:** Given a reference video with wanted semantics as a video prompt, Video-As-Prompt animate a reference image with the same semantics as the reference video.
<p align="center">
<video
controls
autoplay
playsinline
muted
loop
src="https://github.com/user-attachments/assets/2e440927-5b16-4761-ad1f-46ac93de2d8e"
width="60%"
>
Your browser does not support HTML5 video. Here is a <a href="https://github.com/user-attachments/assets/2e440927-5b16-4761-ad1f-46ac93de2d8e">link to the video</a> instead.
</video>
<br>
<em>E.g., Different Reference Videos + Same Reference Image → New Videos with Different Semantics</em>
</p>

> **Welcome to see our [project page](https://bytedance.github.io/Video-As-Prompt) for more interesting results!**

## **Dataset Info**

This is a rich semantic-controlled video generation dataset containing **100 types of control semantics** across **four categories** — **concept**, **style**, **motion**, and **camera**. Each semantic label corresponds to **multiple video clips**, and these clips consistently reflect that specific condition.

This dataset can be used for a wide range of **controllable video generation tasks**, including **in-context video generation**, **video editing**, **visual effects (VFX)**, and more. The original collection contains **over 100,000 video clips**, and after careful filtering and quality control, we publicly release **over 90,000 high-quality clips**.

We warmly welcome the community to explore, research, and build upon this dataset. If you find it helpful, please consider giving us an **upvote ❤️** — your support means a lot!

## 🔗 BibTeX

❤️ If you found this repository helpful, please give us a star and cite our report:

```bibtex
@article{bian2025videoasprompt,
title = {Video-As-Prompt: Unified Semantic Control for Video Generation},
author = {Yuxuan Bian and Xin Chen and Zenan Li and Tiancheng Zhi and Shen Sang and Linjie Luo and Qiang Xu},
journal = {arXiv preprint arXiv:2510.20888},
year = {2025},
url = {https://arxiv.org/abs/2510.20888}
}
```

<div align="center">

# 视频即提示：面向视频生成的统一语义控制

</div>

<div align="center">
<a href=https://bytedance.github.io/Video-As-Prompt target="_blank"><img src=https://img.shields.io/badge/%E9%A1%B9%E7%9B%AE%E4%B8%BB%E9%A1%B5-333399.svg?logo=homepage height=22px></a>
<a href=https://huggingface.co/collections/ByteDance/video-as-prompt target="_blank"><img src=https://img.shields.io/badge/%F0%9F%A4%97%20%E6%A8%A1%E5%9E%8B%E8%B5%84%E6%BA%90-d96902.svg height=22px></a>
<a href=https://huggingface.co/datasets/BianYx/VAP-Data target="_blank"><img src=https://img.shields.io/badge/%F0%9F%A4%97%20%E6%95%B0%E6%8D%AE%E9%9B%86-276cb4.svg height=22px></a>
<a href=https://github.com/bytedance/Video-As-Prompt target="_blank"><img src= https://img.shields.io/badge/%E4%BB%A3%E7%A0%81%E4%BB%93%E5%BA%93-black.svg?logo=github height=22px></a>
<a href=https://arxiv.org/pdf/2510.20888 target="_blank"><img src=https://img.shields.io/badge/%E5%AD%A6%E6%9C%AF%E9%A2%84%E5%8D%B0%E6%9C%AC-b5212f.svg?logo=arxiv height=22px></a>
<!-- <a href=https://yxbian23.github.io/ target="_blank"><img src=https://img.shields.io/badge/Twitter-grey.svg?logo=x height=22px></a> -->
<!-- <a href="https://opensource.org/licenses/Apache">
<img src="https://img.shields.io/badge/License-Apache%202.0-lightgray">
</a> -->
<a href="https://www.youtube.com/watch?v=S3zpLIMOU4c" target="_blank">
<img src="https://img.shields.io/badge/%E2%96%B6%20YouTube%20%E6%BC%94%E7%A4%BA%E8%A7%86%E9%A2%91-FF0000.svg?logo=youtube&logoColor=white" height="24px">
</a>
</div>

<br>

## 🔥 最新动态

- 2025年10月24日：📖 我们发布了首款统一语义控制视频生成模型**Video-As-Prompt（VAP）**！
- 2025年10月24日：🤗 我们发布了**VAP-Data**数据集——当前规模最大的语义可控视频生成数据集，样本量超10万条！
- 2025年10月24日：👋 我们公开了Video-As-Prompt的技术报告，欢迎查阅详情并展开交流探讨！

## 🖌️ **视频即提示（VAP）**

> **核心思路：** 以携带目标语义的参考视频作为视频提示，Video-As-Prompt可依据该参考视频的语义特征，驱动单张参考图像生成动态视频。
<p align="center">
<video
controls
autoplay
playsinline
muted
loop
src="https://github.com/user-attachments/assets/2e440927-5b16-4761-ad1f-46ac93de2d8e"
width="60%"
>
您的浏览器不支持HTML5视频播放，您可以点击<a href="https://github.com/user-attachments/assets/2e440927-5b16-4761-ad1f-46ac93de2d8e">此处查看视频</a>。
</video>
<br>
<em>示例：不同参考视频 + 同一张参考图像 → 生成语义各异的全新视频</em>
</p>

> **欢迎访问我们的[项目主页](https://bytedance.github.io/Video-As-Prompt)查看更多精彩效果！**

## 🔍 数据集说明

本数据集为高质量语义可控视频生成数据集，涵盖**概念、风格、运动、拍摄**四大类别，共计100种控制语义。每种语义标签对应多条视频片段，且所有片段均精准匹配对应语义条件。

本数据集可广泛应用于各类可控视频生成任务，包括**上下文引导视频生成、视频编辑、视觉特效（VFX）**等。原始数据集共收录超10万条视频片段，经严格筛选与质量管控后，本次公开发布的高质量片段超9万条。

我们热忱欢迎学界与社区同仁基于本数据集开展探索与研究工作。若本数据集对您的工作有所助益，欢迎为我们点赞❤️——您的支持对我们意义重大！

## 🔗 引用格式

❤️ 若本仓库对您有所帮助，欢迎点亮Star并引用我们的报告：

bibtex
@article{bian2025videoasprompt,
title = {Video-As-Prompt: Unified Semantic Control for Video Generation},
author = {Yuxuan Bian and Xin Chen and Zenan Li and Tiancheng Zhi and Shen Sang and Linjie Luo and Qiang Xu},
journal = {arXiv preprint arXiv:2510.20888},
year = {2025},
url = {https://arxiv.org/abs/2510.20888}
}