Research map / Vision-language and multimodal / Human and embodied modeling
MotionMaster: Generalizable Text-Driven Motion Generation and Editing
Nan Jiang, Yunhao Li, Lexi Pang, Zimo He, Siyuan Huang, Yixin Zhu
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 |
|---|---|---|---|
| scale training data from 50% to full (ablation) | MotionMaster · single motion generation | Semantic score: 7.90 -> 9.88 | Improved |
| scale training data from 50% to full (ablation) | MotionMaster · long sequence generation | Semantic score: 5.52 -> 7.50 | Improved |
| use default model instead of 3B model (ablation) | MotionMaster · single motion generation | Semantic score: 8.78 -> 9.88 | Improved |
| use default model instead of 3B model (ablation) | MotionMaster · single motion generation | R-Precision R@1: 0.31 -> 0.24 | Hurt |
| add semantic balancing (ablation) | MotionMaster · single motion generation | Semantic score: 8.58 -> 9.88 | Improved |
| add semantic balancing (ablation) | MotionMaster · single motion generation | R-Precision R@1: 0.27 -> 0.24 | Hurt |
| joint training on generation and editing vs generation only (ablation) | MotionMaster · single motion generation | Semantic score: 7.40 -> 9.88 | Improved |
| joint training vs generation only (ablation) | MotionMaster · single motion generation | Diversity: 2.19 -> 1.62 | Hurt |
| joint training vs editing only (ablation) | MotionMaster · motion editing | Semantic score: 8.30 -> 9.10 | Improved |
| joint training vs editing only (ablation) | MotionMaster · motion editing | R-Precision R@1: 0.78 -> 0.77 | No clear effect |
| MLLM-based MotionMaster vs retrained baselines | MMM/MotionLab/MotionMillion · single motion generation | Diversity: 3.10 -> 1.62 | Hurt |
| new FSQ tokenizer with global loss vs MoMask | tokenizer · MotionGB-test | local rotation error (deg): 3.77 -> 7.55 | Hurt |
Closest papers by meaning
- VersatileMotion: A Unified Framework for Motion Synthesis and Comprehension (ECCV 2026 · similarity 0.91)
- MMM: Generative Masked Motion Model (CVPR 2024 · similarity 0.91)
- SnapMoGen: Human Motion Generation from Expressive Texts (NeurIPS 2025 · similarity 0.90)
- Motion-Agent: A Conversational Framework for Human Motion Generation with LLMs (ICLR 2025 · similarity 0.89)
- Dynamic Motion Blending for Versatile Motion Editing (CVPR 2025 · similarity 0.89)
- Scaling Large Motion Models with Million-Level Human Motions (ICML 2025 · similarity 0.89)
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.
