Research map / Video / Video generation
In-Context Generation with Regional Constraints for Instructional Video Editing
Zhongwei Zhang, Fuchen Long, Wei Li, Zhaofan Qiu, Wu Liu, Ting Yao, Tao Mei
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add latent-space regional constraint (ablation) | ReCo · ReCo benchmark (Replace) | edit accuracy SEA: 9.01 -> 9.43 | Improved |
| add latent-space regional constraint (ablation) | ReCo · ReCo benchmark (Add) | edit accuracy SEA: 8.05 -> 8.54 | Improved |
| add latent-space regional constraint (ablation) | ReCo · ReCo benchmark (Remove) | edit accuracy SEA: 6.90 -> 7.28 | Improved |
| add latent-space regional constraint (ablation) | ReCo · ReCo benchmark (Style) | edit accuracy SEA: 9.09 -> 9.43 | Improved |
| add attention-space regional constraint (ablation) | ReCo · ReCo benchmark (Add) | video naturalness SVN: 7.37 -> 7.55 | Improved |
| add attention-space regional constraint (ablation) | ReCo · ReCo benchmark (Replace) | video naturalness SVN: 7.94 -> 8.01 | Improved |
| add attention-space regional constraint (ablation) | ReCo · ReCo benchmark (Remove) | video naturalness SVN: 6.75 -> 6.90 | Improved |
| add attention-space regional constraint (ablation) | ReCo · ReCo benchmark (Style) | overall score S: 9.03 -> 9.17 | Improved |
| use in-context generation instead of ControlNet conditioning (ablation) | ReCo · ReCo benchmark (Replace) | edit accuracy SEA: 8.01 -> 9.43 | Improved |
| use in-context generation instead of ControlNet conditioning (ablation) | ReCo · ReCo benchmark (Add) | overall score S: 7.83 -> 8.23 | Improved |
| use in-context generation instead of ControlNet conditioning (ablation) | ReCo · ReCo benchmark (Remove) | overall score S: 6.58 -> 7.00 | Improved |
| use in-context generation instead of ControlNet conditioning (ablation) | ReCo · ReCo benchmark (Style) | overall score S: 8.93 -> 9.17 | Improved |
Closest papers by meaning
- Señorita-2M: A High-Quality Instruction-based Dataset for General Video Editing by Video Specialists (NeurIPS 2025 · similarity 0.87)
- VicEdit: Learning to Edit Videos from Visual In-Context Examples (NeurIPS 2026 · similarity 0.87)
- VINCIE-NExT: Unlocking Video Editing from Images via In-Context Modeling (NeurIPS 2026 · similarity 0.86)
- VEU-Bench: Towards Comprehensive Understanding of Video Editing (CVPR 2025 · similarity 0.85)
- VideoDirector: Precise Video Editing via Text-to-Video Models (CVPR 2025 · similarity 0.85)
- Enabling Instructional Image Editing with In-Context Generation in Large Scale Diffusion Transformer (NeurIPS 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
