Research map / Video / Video generation
NOVA: Sparse Control, Dense Synthesis for Pair-Free Video Editing
Tianlin Pan, Jiayi Dai, Chenpu Yuan, Zhengyao Lv, Binxin Yang, Hubery Yin, Chen Li, Jing Lyu, Caifeng Shan, Chenyang Si
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 multi-keyframe guidance instead of first frame only | VACE 1.3B · Pexels test videos | Success rate: 0.36 -> 0.90 | Improved |
| add multi-keyframe guidance | LoRA-Edit 14B · Pexels test videos | Success rate: 0.80 -> 0.86 | Improved |
| use NOVA dual-branch sparse/dense design instead of multi-keyframe VACE | VACE 1.3B · Pexels test videos | FC: 0.840 -> 0.882 | Improved |
| use NOVA instead of multi-keyframe VACE | VACE 1.3B · Pexels test videos | BG-SSIM: 0.913 -> 0.917 | No clear effect |
| use NOVA instead of Senorita-2M | Senorita-2M 5B · Pexels test videos | VBench Background Consistency: 0.953 -> 0.946 | Hurt |
| add dense branch (ablation) | NOVA · Object addition/removal GT pairs | CLIP similarity to ground truth | Improved |
| anchor later keyframe edits to the first edited frame (ablation) | NOVA inference · Pexels test videos | visual style consistency | Improved |
| replace FLUX.1 Kontext with Qwen-Image-Edit-2509 as keyframe editor (ablation) | NOVA · Pexels test videos | TC: 0.92 -> 0.88 | Hurt |
| replace FLUX.1 Kontext with Qwen-Image-Edit-2509 as keyframe editor (ablation) | NOVA · Pexels test videos | FC: 0.85 -> 0.87 | Improved |
| replace FLUX.1 Kontext with Qwen-Image-Edit-2509 as keyframe editor (ablation) | NOVA · Pexels test videos | BG-SSIM: 0.88 -> 0.89 | Improved |
| vary inference keyframe interval to 8, 16, 20 versus training interval 10 (ablation) | NOVA · Pexels test videos | editing quality | No clear effect |
Closest papers by meaning
- MotionV2V: Editing Motion in a Video (CVPR 2026 · similarity 0.89)
- VINCIE-NExT: Unlocking Video Editing from Images via In-Context Modeling (NeurIPS 2026 · similarity 0.88)
- Transforming Image Editors into Video Editors (NeurIPS 2026 · similarity 0.88)
- MotionEditor: Editing Video Motion via Content-Aware Diffusion (CVPR 2024 · similarity 0.87)
- Fairy: Fast Parallelized Instruction-Guided Video-to-Video Synthesis (CVPR 2024 · similarity 0.87)
- Align-A-Video: Deterministic Reward Tuning of Image Diffusion Models for Consistent Video Editing (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.
