Research map / Video / Video generation
MoSA: Motion-Coherent Human Video Generation via Structure-Appearance Decoupling
Haoyu Wang, Hao Tang, Donglin Di, Zhilu Zhang, Wangmeng Zuo, Feng Gao, Siwei Ma, Shiliang Zhang
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 structure-appearance decoupling (3D structure branch) to the finetuned base model (ablation) | CogVideoX-5B · MoVid human video generation | motion quality / FVD | Improved |
| use 3D structure generator instead of 2D skeleton generation model (ablation) | MoSA · MoVid human video generation | structure correctness | Improved |
| add HADC modules (ablation) | MoSA · MoVid human video generation | motion consistency | Improved |
| add motion loss L_m (ablation) | MoSA · MoVid human video generation | motion realism | Improved |
| add dense tracking loss with interval-dependent weights over static weights (ablation) | MoSA · MoVid human video generation | temporal motion consistency | Improved |
| add contact constraint L_cont (ablation) | MoSA · MoVid human video generation | interpenetration artifacts | Improved |
| apply MoSA decoupling framework to Wan 2.1 backbone (ablation) | Wan 2.1 · MoVid human video generation | video quality | Improved |
| use MoSA over Wan 2.1 | Wan 2.1 -> MoSA · Text-to-video comparison | FVD: 1251 -> 1093 | Improved |
| use MoSA over Uni3C on HumanVid | Uni3C · HumanVid dataset pose-driven | FVD: 562.7 -> 568.3 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- VideoCrafter2 — VideoCrafter2: Overcoming Data Limitations for High-Quality Video Diffusion Models (CVPR 2024) · better on 2 of 2 shared results
- LaVie — LaVie: High-Quality Video Generation with Cascaded Latent Diffusion Models (IJCV 2024) · 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
- The Quest for Generalizable Motion Generation: Data, Model, and Evaluation (ICLR 2026 · similarity 0.89)
- HumanDreamer: Generating Controllable Human-Motion Videos via Decoupled Generation (CVPR 2025 · similarity 0.89)
- HumanVid: Demystifying Training Data for Camera-controllable Human Image Animation (NeurIPS 2024 · similarity 0.88)
- OpenHumanVid: A Large-Scale High-Quality Dataset for Enhancing Human-Centric Video Generation (CVPR 2025 · similarity 0.87)
- VideoJAM: Joint Appearance-Motion Representations for Enhanced Motion Generation in Video Models (ICML 2025 · similarity 0.87)
- ONE-SHOT: Compositional Human-Environment Video Synthesis via Spatial-Decoupled Motion Injection and Hybrid Context Integration (NeurIPS 2026 · 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.
