Research map / Video / Video generation
PAI-Studio: Cinematic Video Background Replacement with Camera-Aware Motion
Heyuan Gao, Bangxun Tang, Yiren Song, Guian Fang, Zijian He, Jie Yang, Mike Zheng Shou
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 temporal positional encoding cloning (ablation) | PAI-Studio · Cine-Restore (single-frame) | Motion consistency: 0.824 -> 0.872 | Improved |
| add temporal positional encoding cloning (ablation) | PAI-Studio · Cine-Restore (single-frame) | MSE: 0.0300 -> 0.0164 | Improved |
| add temporal positional encoding cloning (ablation) | PAI-Studio · Cine-Restore (multi-frame) | Motion consistency: 0.843 -> 0.894 | Improved |
| add illumination perturbation augmentation (ablation) | PAI-Studio · Cine-Restore (single-frame) | Illumination harmony: 6.44 -> 9.11 | Improved |
| add illumination perturbation augmentation (ablation) | PAI-Studio · Cine-Restore (multi-frame) | Illumination harmony: 7.31 -> 9.61 | Improved |
| add illumination perturbation augmentation (ablation) | PAI-Studio · Cine-Restore (single-frame) | Motion consistency: 0.850 -> 0.872; MSE 0.0270 -> 0.0164 | Mixed |
| use structured annotations instead of simple text prompts (ablation) | PAI-Studio · Cine-Restore (single-frame) | MSE: 0.0177 -> 0.0164 | Improved |
| use structured annotations instead of simple text prompts (ablation) | PAI-Studio · Cine-Restore (single-frame) | Illumination harmony: 8.88 -> 9.11 | Improved |
| add multi-frame background control (3 references) versus single frame | PAI-Studio · Cine-Restore | MSE: 0.0164 -> 0.00645 | Improved |
| add multi-frame background control versus single frame | PAI-Studio · Cine-Restore | LPIPS: 0.303 -> 0.197 | Improved |
| add multi-frame background control versus single frame | Kling 3.0 Omni · Cine-Restore | Motion consistency: 0.787 -> 0.715 | Hurt |
Closest papers by meaning
- PAI-Actor: Cinematic Multi-Actor Character Replacement in Dynamic Scenes (NeurIPS 2026 · similarity 0.89)
- Sparkle: Realizing Lively Instruction-Guided Video Background Replacement via Decoupled Guidance (NeurIPS 2026 · similarity 0.87)
- Frame In-N-Out: Unbounded Controllable Image-to-Video Generation (NeurIPS 2025 · similarity 0.87)
- IllumiCraft: Unified Geometry and Illumination Diffusion for Controllable Video Generation (NeurIPS 2025 · similarity 0.86)
- ActAnywhere: Subject-Aware Video Background Generation (NeurIPS 2024 · similarity 0.86)
- FlowVid: Taming Imperfect Optical Flows for Consistent Video-to-Video Synthesis (CVPR 2024 · 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.
