VideoCoF: Unified Video Editing with Temporal Reasoner
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908938333585408 |
|---|---|
| author | Yang, Xiangpeng Xie, Ji Yang, Yiyuan Ma, Yue Huang, Yan Xu, Min Wu, Qiang |
| author_facet | Yang, Xiangpeng Xie, Ji Yang, Yiyuan Ma, Yue Huang, Yan Xu, Min Wu, Qiang |
| contents | Existing video editing methods face a critical trade-off: expert models offer precision but rely on task-specific priors like masks, hindering unification; conversely, unified temporal in-context learning models are mask-free but lack explicit spatial cues, leading to weak instruction-to-region mapping and imprecise localization. To resolve this conflict, we propose VideoCoF, a novel Chain-of-Frames approach inspired by Chain-of-Thought reasoning. VideoCoF enforces a ``see, reason, then edit" procedure by compelling the video diffusion model to first predict reasoning tokens (edit-region latents) before generating the target video tokens. This explicit reasoning step removes the need for user-provided masks while achieving precise instruction-to-region alignment and fine-grained video editing. Furthermore, we introduce a RoPE alignment strategy that leverages these reasoning tokens to ensure motion alignment and enable length extrapolation beyond the training duration. We demonstrate that with a minimal data cost of only 50k video pairs, VideoCoF achieves state-of-the-art performance on VideoCoF-Bench, validating the efficiency and effectiveness of our approach. Our code, weight, data are available at https://github.com/knightyxp/VideoCoF. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_07469 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | VideoCoF: Unified Video Editing with Temporal Reasoner Yang, Xiangpeng Xie, Ji Yang, Yiyuan Ma, Yue Huang, Yan Xu, Min Wu, Qiang Computer Vision and Pattern Recognition Existing video editing methods face a critical trade-off: expert models offer precision but rely on task-specific priors like masks, hindering unification; conversely, unified temporal in-context learning models are mask-free but lack explicit spatial cues, leading to weak instruction-to-region mapping and imprecise localization. To resolve this conflict, we propose VideoCoF, a novel Chain-of-Frames approach inspired by Chain-of-Thought reasoning. VideoCoF enforces a ``see, reason, then edit" procedure by compelling the video diffusion model to first predict reasoning tokens (edit-region latents) before generating the target video tokens. This explicit reasoning step removes the need for user-provided masks while achieving precise instruction-to-region alignment and fine-grained video editing. Furthermore, we introduce a RoPE alignment strategy that leverages these reasoning tokens to ensure motion alignment and enable length extrapolation beyond the training duration. We demonstrate that with a minimal data cost of only 50k video pairs, VideoCoF achieves state-of-the-art performance on VideoCoF-Bench, validating the efficiency and effectiveness of our approach. Our code, weight, data are available at https://github.com/knightyxp/VideoCoF. |
| title | VideoCoF: Unified Video Editing with Temporal Reasoner |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2512.07469 |