Research map / Video / Video generation
HorizonRelight: Relighting Long-horizon Videos Consistently via Diffusion Transformers
Jing Yang, Mayoore Jaiswal, Zian Wang, Xiao Zeng, Yajie Zhao, Jianyuan Min, Rochelle Pereira
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 chained propagated target-domain context with masked self-conditioning instead of independent chunks | DiffusionRenderer · synthetic boundary MSE | relighting MSE: 0.1048 -> 0.0729 | Improved |
| use chained propagated context instead of independent chunks | DiffusionRenderer · synthetic boundary MSE | metallic MSE: 0.2050 -> 0.0542 | Improved |
| use chained propagated context instead of independent chunks | DiffusionRenderer · synthetic boundary MSE | roughness MSE: 0.2614 -> 0.1042 | Improved |
| use chained propagated context instead of independent chunks | DiffusionRenderer · synthetic sequence MSE | depth MSE: 0.2208 -> 0.1383 | Improved |
| use chained propagated context instead of independent chunks | DiffusionRenderer · synthetic sequence MSE | relighting MSE: 0.1063 -> 0.0906 | Improved |
| use chained propagated context instead of independent chunks | DiffusionRenderer · synthetic boundary MSE | specular MSE: 0.1205 -> 0.0935 | Improved |
| disable warm-start G-buffer for a misaligned prompted frame (ablation) | warm-start prompting · in-the-wild YouTube CC (qualitative) | visual quality | Improved |
| add warm-start image prompt over cold start (ablation) | HorizonRelight · in-the-wild (qualitative) | specular detail | Improved |
| use video prompting instead of image prompting for warm start (ablation) | warm-start prompting · in-the-wild (qualitative) | robustness | Hurt |
| extend chain beyond about five chunks (300 frames) | HorizonRelight · long-horizon limitation | fine detail | Hurt |
Closest papers by meaning
- UniRelight: Learning Joint Decomposition and Synthesis for Video Relighting (NeurIPS 2025 · similarity 0.90)
- Lux Post Facto: Learning Portrait Performance Relighting with Conditional Video Diffusion and a Hybrid Dataset (CVPR 2025 · similarity 0.88)
- UniLumos: Fast and Unified Image and Video Relighting with Physics-Plausible Feedback (NeurIPS 2025 · similarity 0.88)
- VSTAR: Generative Temporal Nursing for Longer Dynamic Video Synthesis (ICLR 2025 · similarity 0.86)
- IllumiCraft: Unified Geometry and Illumination Diffusion for Controllable Video Generation (NeurIPS 2025 · similarity 0.86)
- Re-ttention: Ultra Sparse Visual Generation via Attention Statistical Reshape (NeurIPS 2025 · similarity 0.86)
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.
