Research map / Video / Video generation
Go to Zero: Towards Zero-shot Motion Generation with Million-scale Data
Ke Fan, Shunlin Lu, Minyue Dai, Runyi Yu, Lixing Xiao, Zhiyang Dou, Junting Dong, Lizhuang Ma, Jingbo Wang
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 wavelet transform before FSQ encoder and inverse after decoder (ablation) | FSQ motion tokenizer · MotionMillion | MPJPE: 46.8 -> 45.5 | Improved |
| add wavelet transform to FSQ (ablation) | FSQ motion tokenizer · MotionMillion | Mean acceleration deviation: 6.0 -> 4.0 | Improved |
| add wavelet transform to FSQ (ablation) | FSQ motion tokenizer · MotionMillion | Max acceleration: 15.0 -> 12.0 | Improved |
| train FSQ on MotionMillion instead of MotionUnion (ScaMo) | FSQ tokenizer · HumanML3D | MPJPE: 63.3 -> 41.9 | Improved |
| train FSQ on MotionMillion instead of MotionUnion (ScaMo) | FSQ tokenizer · MotionMillion | MPJPE: 88.9 -> 45.5 | Improved |
| scale model from 1B to 3B | Ours text-to-motion · MotionMillion test set | FID: 31.3 -> 10.8 | Improved |
| scale model from 3B to 7B | Ours text-to-motion · MotionMillion test set | FID: 10.8 -> 10.3 | No clear effect |
| scale model from 1B to 7B | Ours text-to-motion · MotionMillion-Eval | Text Alignment: 170.3 -> 261 | Improved |
| scale model from 1B to 7B | Ours text-to-motion · MotionMillion-Eval | Motion Smoothness: 501 -> 501 | No clear effect |
| scale model from 1B to 7B | Ours text-to-motion · MotionMillion-Eval | Physical Plausibility: 497 -> 495.5 | No clear effect |
| use MotionMillion-trained 3B model instead of ScaMo-3B | text-to-motion · MotionMillion-Eval | Text Alignment: 226.6 -> 238.6 | Improved |
| use MotionMillion-trained 3B model instead of ScaMo-3B | text-to-motion · MotionMillion-Eval | Physical Plausibility: 477.5 -> 496 | Improved |
Closest papers by meaning
- Scaling Large Motion Models with Million-Level Human Motions (ICML 2025 · similarity 0.90)
- OpenT2M: No-frill Motion Generation with Open-source, Large-scale, High-quality Data (CVPR 2026 · similarity 0.90)
- FineMotion: A Dataset and Benchmark with both Spatial and Temporal Annotation for Fine-grained Motion Generation and Editing (ICCV 2025 · similarity 0.89)
- HumanVid: Demystifying Training Data for Camera-controllable Human Image Animation (NeurIPS 2024 · similarity 0.88)
- MMM: Generative Masked Motion Model (CVPR 2024 · similarity 0.87)
- HumanDreamer: Generating Controllable Human-Motion Videos via Decoupled Generation (CVPR 2025 · 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.
