Research map / Video / Video generation
EasyV2V: A High-quality Instruction-based Video Editing Framework
Jinjie Mai, Chaoyang Wang, Gordon Guocheng Qian, Willi Menapace, Sergey Tulyakov, Bernard Ghanem, Peter Wonka, Ashkan Mirzaei
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 |
|---|---|---|---|
| use sequence concatenation instead of embedding addition for source video (ablation) | EasyV2V with LoRA · EditVerse architecture ablation | VLM score @ 40K: 6.29 -> 7.47 | Improved |
| use LoRA instead of full finetuning (ablation) | EasyV2V with SeqCat · EditVerse architecture ablation | VLM score @ 40K: 3.94 -> 7.47 | Improved |
| use LoRA instead of full finetuning (ablation) | EasyV2V with EmbedAdd · EditVerse architecture ablation | VLM score @ 40K: 4.57 -> 6.29 | Improved |
| add affine-transformed I2I pairs as pseudo videos to single-image I2I (ablation) | V2V editor · I2I data ablation | VLM edit quality: 5.52 -> 6.24 | Improved |
| add I2I data (single + affine) to V2V training data (ablation) | V2V editor · I2I data ablation | VLM edit quality: 6.69 -> 6.86 | Improved |
| train on Human Animate vs Actor Transmutation (Flow Edit) data (ablation) | V2V editor · Dataset ablation, Actor transmutation | VLM score: 7.20 -> 8.30 | Hurt |
| add dense-caption data for human action edits (ablation) | V2V editor · Dataset ablation, Change human action | VLM score: 4.97 -> 6.87 | Improved |
| add stylization data (ablation) | V2V editor · Dataset ablation, Stylization | VLM score: 4.97 -> 7.97 | Improved |
| inject mask by addition into source latent instead of Wan VACE branch (ablation) | mask control · Mask conditioning comparison | VLM score (average): 5.50 -> 7.48 | Improved |
| inject mask into source latent instead of target latent (ablation) | mask control · Mask conditioning comparison | VLM score (average): 6.45 -> 7.48 | Improved |
| raise LoRA rank from 64 to 256 (ablation) | EasyV2V · EditVerse LoRA rank | VLM quality score (w/o Ref.): 6.17 -> 6.48 | Improved |
| add reference image from an image editor (ablation) | EasyV2V rank 256 · EditVerse LoRA rank | VLM quality score: 6.48 -> 7.22 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- EditVerse — EditVerse: Unifying Image and Video Editing and Generation with In-Context Learning (ICLR 2026) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Transforming Image Editors into Video Editors (NeurIPS 2026 · similarity 0.89)
- VINCIE-NExT: Unlocking Video Editing from Images via In-Context Modeling (NeurIPS 2026 · similarity 0.89)
- Controllable First-Frame-Guided Video Editing via Mask-Aware LoRA Fine-Tuning (ICLR 2026 · similarity 0.89)
- Señorita-2M: A High-Quality Instruction-based Dataset for General Video Editing by Video Specialists (NeurIPS 2025 · similarity 0.88)
- VicEdit: Learning to Edit Videos from Visual In-Context Examples (NeurIPS 2026 · similarity 0.87)
- VideoDirector: Precise Video Editing via Text-to-Video Models (CVPR 2025 · similarity 0.87)
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.
