---
license: apache-2.0
library_name: vincie
pipeline_tag: image-to-image
---

<!-- <div align="center">
  <img src="assets/seedvr_logo.png" alt="SeedVR" width="400"/>
</div> -->


# VINCIE: Unlocking In-context Image Editing from Video
> [Leigang Qu](https://leigang-qu.github.io/), [Feng Cheng](https://klauscc.github.io/), [Ziyan Yang](https://ziyanyang.github.io/), [Qi Zhao](https://kevinz8866.github.io/), [Shanchuan Lin](https://scholar.google.com/citations?user=EDWUw7gAAAAJ&hl=en), [Yichun Shi](https://seasonsh.github.io/), [Yicong Li](https://yl3800.github.io/), [Wenjie Wang](https://wenjiewwj.github.io/), [Tat-Seng Chua](https://www.chuatatseng.com/), [Lu Jiang](http://www.lujiang.info/index.html)

<p align="center">
  <a href="https://vincie2025.github.io/">
    <img
      src="https://img.shields.io/badge/VINCIE-Website-0A66C2?logo=safari&logoColor=white"
      alt="VINCIE Website"
    />
  </a>
  <a href="https://arxiv.org/abs/2506.10941">
    <img
      src="https://img.shields.io/badge/VINCIE-Paper-red?logo=arxiv&logoColor=red"
      alt="VINCIE Paper on ArXiv"
    />
  </a>
  <a href="https://github.com/ByteDance-Seed/VINCIE">
            <img 
              alt="Github" src="https://img.shields.io/badge/VINCIE-Codebase-536af5?color=536af5&logo=github"
              alt="VINCIE Codebase"
            />
  </a>
  <a href="https://huggingface.co/collections/ByteDance-Seed/vincie-6864cc2e3116d82e4a83a17c">
    <img 
        src="https://img.shields.io/badge/VINCIE-Models-yellow?logo=huggingface&logoColor=yellow" 
        alt="VINCIE Models"
    />
  </a>
   <a href="https://huggingface.co/spaces/ByteDance-Seed/VINCIE-3B">
    <img 
        src="https://img.shields.io/badge/VINCIE-Space-orange?logo=huggingface&logoColor=yellow" 
        alt="VINCIE Space"
    />
  </a>
</p>

>
> In-context image editing aims to modify images based on a contextual sequence comprising text and previously generated images. Existing methods typically depend on task-specific pipelines and expert models (e.g., segmentation and inpainting) to curate training data. In this work, we explore whether an in-context image editing model can be learned directly from videos. We introduce a scalable approach to annotate videos as interleaved multimodal sequences. To effectively learn from this data, we design a block-causal diffusion transformer trained on three proxy tasks: next-image prediction, current segmentation prediction, and next-segmentation prediction. Additionally, we propose a novel multi-turn image editing benchmark to advance research in this area. Extensive experiments demonstrate that our model exhibits strong in-context image editing capabilities and achieves state-of-the-art results on two multi-turn image editing benchmarks. Despite being trained exclusively on videos, our model also shows promising abilities in multi-concept composition, story generation, and chain-of-editing applications.



## ✍️ Citation

```bibtex
@article{qu2025vincie,
  title={VINCIE: Unlocking In-context Image Editing from Video},
  author={Qu, Leigang and Cheng, Feng and Yang, Ziyan and Zhao, Qi and Lin, Shanchuan and Shi, Yichun and Li, Yicong and Wang, Wenjie and Chua, Tat-Seng and Jiang, Lu},
  journal={arXiv preprint arXiv:2506.10941},
  year={2025}
}

```


## 📜 License
VINCIE is licensed under the Apache 2.0.