Research map / Video / Video generation
OpenT2M: No-frill Motion Generation with Open-source, Large-scale, High-quality Data
Bin Cao, Sipeng Zheng, Hao Luo, Boyuan Li, Jing Liu, Zongqing Lu
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 |
|---|---|---|---|
| train on OpenT2M instead of HumanML3D | MonoFrill-2D-PRQ4 · OpenT2M_zero (zero-shot) | FID: 60.177 -> 1.475 | Improved |
| train on OpenT2M instead of HumanML3D | MDM · OpenT2M_zero (zero-shot) | FID: 51.307 -> 8.153 | Improved |
| train on OpenT2M instead of Motion-X | T2M-GPT · OpenT2M_zero (zero-shot) | R@1: 0.063 -> 0.159 | Improved |
| train on OpenT2M instead of Motion-X | Being-M0 · OpenT2M_zero (zero-shot) | FID: 46.222 -> 5.811 | Improved |
| add OpenT2M pretraining (ablation) | MonoFrill GPT2-medium · HumanML3D (fine-tuned 50 epochs) | FID: 61.809 -> 17.91 | Improved |
| add OpenT2M pretraining (ablation) | MonoFrill LlaMA3-8B · HumanML3D (fine-tuned 50 epochs) | FID: 0.546 -> 0.238 | Improved |
| add OpenT2M pretraining (ablation) | MonoFrill LlaMA2-7B · HumanML3D (fine-tuned 50 epochs) | R@1: 0.472 -> 0.485 | Improved |
| add long-horizon motion data to training (ablation) | MonoFrill · OpenT2M_long | FID: 36.837 -> 0.430 | Improved |
| add text refinement (ablation) | MonoFrill + long-horizon data · OpenT2M_long | FID: 0.430 -> 0.297 | Improved |
| add text refinement (ablation) | MonoFrill · HumanML3D | R@1: 0.520 -> 0.533 | Improved |
| replace PRQ4 with 2D-PRQ4 tokenizer | MonoFrill tokenizer · OpenT2M reconstruction | MPJPE: 95.743 -> 49.134 | Improved |
| replace PRQ6 with 2D-PRQ6 tokenizer | tokenizer · HumanML3D reconstruction | MPJPE: 25.485 -> 25.417 | No clear effect |
Closest papers by meaning
- Open the Motion Door: Atomic Motion Decomposition and Recomposition for Open-Vocabulary Motion Generation (CVPR 2026 · similarity 0.92)
- TM2T: Stochastic and Tokenized Modeling for the Reciprocal Generation of 3D Human Motions and Texts (ECCV 2022 · similarity 0.91)
- Light-T2M: A Lightweight and Fast Model for Text-to-motion Generation (AAAI 2025 · similarity 0.91)
- Scaling Large Motion Models with Million-Level Human Motions (ICML 2025 · similarity 0.91)
- MMM: Generative Masked Motion Model (CVPR 2024 · similarity 0.90)
- SnapMoGen: Human Motion Generation from Expressive Texts (NeurIPS 2025 · similarity 0.90)
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.
