Research map / Video / Video generation
FLAME: Free-Form Language-Based Motion Synthesis & Editing
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| replace TEMOS with FLAME | text-to-motion generation · HumanML3D | mCLIP: 0.254 -> 0.297 | Helped |
| replace TEMOS with FLAME | text-to-motion generation · HumanML3D | FD: 49.142 -> 21.152 | Helped |
| replace TEMOS with FLAME | text-to-motion generation · BABEL | R-Precision Top-1: 0.786 -> 0.888 | Helped |
| replace TEMOS with FLAME | text-to-motion generation · KIT | AVE mean global: 0.448 -> 0.500 | Hurt |
| replace TEMOS with FLAME | text-to-motion generation · HumanML3D | Joint Diversity: 0.017 -> 0.072 | Helped |
| add motion-length token (ablation) | FLAME self-attention architecture · HumanML3D | mCLIP: 0.239 -> 0.239 | No clear effect |
| add motion-length token (ablation) | FLAME self-attention architecture · HumanML3D | FD: 60.142 -> 59.254 | Helped |
| add cross-attention using full language tokens (ablation) | FLAME with motion-length token · HumanML3D | mCLIP: 0.239 -> 0.290 | Helped |
| add cross-attention using full language tokens (ablation) | FLAME with motion-length token · HumanML3D | FD: 59.254 -> 27.157 | Helped |
| freeze the pretrained language model (ablation) | FLAME with cross-attention · HumanML3D | R-Precision Top-1: 0.439 -> 0.513 | Helped |
| freeze the pretrained language model (ablation) | FLAME with cross-attention · HumanML3D | FD: 27.157 -> 21.152 | Helped |
| reduce sampling steps from 1,000 to 5 (ablation) | FLAME trained with T=1000 · motion sampling | Time Elapsed (s): 32.81 -> 0.72 | Helped |
Closest papers by meaning
- MMM: Generative Masked Motion Model (CVPR 2024 · similarity 0.89)
- Enhanced Fine-Grained Motion Diffusion for Text-Driven Human Motion Synthesis (AAAI 2024 · similarity 0.88)
- SimMotionEdit: Text-Based Human Motion Editing with Motion Similarity Prediction (CVPR 2025 · similarity 0.88)
- SALAD: Skeleton-aware Latent Diffusion for Text-driven Motion Generation and Editing (CVPR 2025 · similarity 0.87)
- AMD: Autoregressive Motion Diffusion (AAAI 2024 · similarity 0.87)
- Dynamic Motion Blending for Versatile Motion Editing (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.