Research map / Video / Video generation
PoseAnything: General Pose-guided Video Generation with Part-aware Temporal Coherence
Ruiyan Wang, Teng Hu, Kaihui Huang, Zihan Su, Ran Yi, Lizhuang Ma
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 Part-aware Temporal Coherence Module over channel-concat baseline (ablation) | PoseAnything · XPose-benchmark | FVD: 102.30 -> 99.970 | Improved |
| use part segmentation and matching instead of whole-object cross-attention (ablation) | PTCM · XPose-benchmark | FVD: 101.50 -> 99.970 | Improved |
| whole-object cross-attention (EC) versus no coherence module (ablation) | PoseAnything · XPose-benchmark | PSNR: 29.85 -> 30.27 | Improved |
| inject pose by channel concatenation rather than MLP (ablation) | PoseAnything (no PTCM) · TikTok | FVD: 283.79 -> 133.95 | Improved |
| inject pose by channel concatenation rather than width concatenation (ablation) | PoseAnything (no PTCM) · TikTok | FVD: 415.13 -> 133.95 | Improved |
| reduce pose injection from 100% to 10% of frames (sparse training) (ablation) | PoseAnything · XPose-benchmark | FVD: 99.97 -> 97.02 | No clear effect |
| reduce pose injection to 2.5% of frames (ablation) | PoseAnything · XPose-benchmark | SSIM: 0.7114 -> 0.6757 | Hurt |
| raise CFG scale from 1.2 to 5.0 under dense pose (ablation) | PoseAnything · XPose-benchmark (dense) | PSNR: 30.61 -> 29.92 | Hurt |
| raise CFG scale from 1.2 to 5.0 under sparse pose (ablation) | PoseAnything · XPose-benchmark (sparse) | PSNR: 30.68 -> 29.84 | Hurt |
| train on XPose non-human data and use skeleton conditioning vs drag-based methods | PoseAnything · XPose-benchmark (non-human) | LPIPS: 0.3530 -> 0.3241 | Improved |
| PoseAnything vs Animate-X on human videos | PoseAnything · TikTok (human) | FVD: 139.01 -> 133.95 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- Animate-X — Animate-X: Universal Character Image Animation with Enhanced Motion Representation (ICLR 2025) · better on 4 of 4 shared results
- Tora — Tora: Trajectory-oriented Diffusion Transformer for Video Generation (CVPR 2025) · better on 2 of 2 shared results
- SG-I2V — SG-I2V: Self-Guided Trajectory Control in Image-to-Video Generation (ICLR 2025) · better on 1 of 1 shared result
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
- HumanDreamer: Generating Controllable Human-Motion Videos via Decoupled Generation (CVPR 2025 · similarity 0.90)
- HumanVid: Demystifying Training Data for Camera-controllable Human Image Animation (NeurIPS 2024 · similarity 0.88)
- DisPose: Disentangling Pose Guidance for Controllable Human Image Animation (ICLR 2025 · similarity 0.88)
- MotionBooth: Motion-Aware Customized Text-to-Video Generation (NeurIPS 2024 · similarity 0.87)
- MimicMotion: High-Quality Human Motion Video Generation with Confidence-aware Pose Guidance (ICML 2025 · similarity 0.87)
- Follow Your Pose: Pose-Guided Text-to-Video Generation Using Pose-Free Videos (AAAI 2024 · 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.
