Research map / Video / Video generation
Light-A-Video: Training-free Video Relighting via Progressive Light Fusion
Yujie Zhou, Jiazi Bu, Pengyang Ling, Pan Zhang, Tong Wu, Qidong Huang, Jinsong Li, Xiaoyi Dong, Yuhang Zang, Yuhang Cao, Anyi Rao, Jiaqi Wang and 1 more
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 |
|---|---|---|---|
| apply Light-A-Video (CLA + PLF with video diffusion prior) instead of per-frame IC-Light | IC-Light · Video relighting test set | temporal CLIP score: 0.9040 -> 0.9667 | Improved |
| apply Light-A-Video instead of per-frame IC-Light | IC-Light · Video relighting test set | motion preservation: 5.969 -> 1.833 | Improved |
| apply Light-A-Video instead of per-frame IC-Light | IC-Light · User study | Video Smoothness: 2.128 -> 3.502 | Improved |
| use Light-A-Video instead of SDEdit with 0.2 noise | IC-Light · Video relighting test set | FID: 13.79 -> 29.63 | Hurt |
| add more noise (SDEdit 0.6) for temporal smoothing | IC-Light + SDEdit · Video relighting test set | ID preservation: FID 13.79 -> 62.61 | Hurt |
| add Consistent Light Attention (CLA) to IC-Light (ablation) | Light-A-Video · qualitative ablation with background generation | lighting source stability | Improved |
| add Progressive Light Fusion (PLF) with video diffusion priors (ablation) | Light-A-Video · qualitative ablation | temporal smoothness of relit appearance | Improved |
| use decaying fusion weight lambda_t in PLF instead of constant lambda_t = 1 (ablation) | PLF · FID, temporal CLIP and motion preservation | FID / motion preservation | Improved |
| increase CLA trade-off gamma to large values (ablation) | CLA · FID versus gamma | FID | Hurt |
| use CLA gamma between 0.2 and 0.5 (ablation) | CLA · FID and motion preservation | FID / motion preservation | Improved |
Later papers that compared with this method
3 later accepted papers report Light-A-Video as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Relightful Video Portrait Harmonization (CVPR 2026 · better on 4 of 4 shared results)
- Light-X: Generative 4D Video Rendering with Camera and Illumination Control (ICLR 2026 · better on 1 of 1 shared result)
- IllumiCraft: Unified Geometry and Illumination Diffusion for Controllable Video Generation (NeurIPS 2025 · better on 8 of 8 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
- Lux Post Facto: Learning Portrait Performance Relighting with Conditional Video Diffusion and a Hybrid Dataset (CVPR 2025 · similarity 0.91)
- Comprehensive Relighting: Generalizable and Consistent Monocular Human Relighting and Harmonization (CVPR 2025 · similarity 0.91)
- UniRelight: Learning Joint Decomposition and Synthesis for Video Relighting (NeurIPS 2025 · similarity 0.91)
- UniLumos: Fast and Unified Image and Video Relighting with Physics-Plausible Feedback (NeurIPS 2025 · similarity 0.91)
- DreamLight: Towards Harmonious and Consistent Image Relighting (NeurIPS 2025 · similarity 0.89)
- Real-time 3D-aware Portrait Video Relighting (CVPR 2024 · similarity 0.88)
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.
