VINCIE: Unlocking In-context Image Editing from Video

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Qu, Leigang, Cheng, Feng, Yang, Ziyan, Zhao, Qi, Lin, Shanchuan, Shi, Yichun, Li, Yicong, Wang, Wenjie, Chua, Tat-Seng, Jiang, Lu
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915825031577600
author Qu, Leigang
Cheng, Feng
Yang, Ziyan
Zhao, Qi
Lin, Shanchuan
Shi, Yichun
Li, Yicong
Wang, Wenjie
Chua, Tat-Seng
Jiang, Lu
author_facet Qu, Leigang
Cheng, Feng
Yang, Ziyan
Zhao, Qi
Lin, Shanchuan
Shi, Yichun
Li, Yicong
Wang, Wenjie
Chua, Tat-Seng
Jiang, Lu
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2506_10941
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle VINCIE: Unlocking In-context Image Editing from Video
Qu, Leigang
Cheng, Feng
Yang, Ziyan
Zhao, Qi
Lin, Shanchuan
Shi, Yichun
Li, Yicong
Wang, Wenjie
Chua, Tat-Seng
Jiang, Lu
Computer Vision and Pattern Recognition
Artificial Intelligence
Computation and Language
Machine Learning
Multimedia
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.
title VINCIE: Unlocking In-context Image Editing from Video
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Computation and Language
Machine Learning
Multimedia
url https://arxiv.org/abs/2506.10941