Research map / Video / Video generation
UDE: A Unified Driving Engine for Human Motion Generation
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 GRU-based UTT with GPT-based UTT (ablation) | UDE · HumanML3D | Top-1 Acc.: 7.89 -> 8.11 | Helped |
| Replace GRU-based UTT with GPT-based UTT (ablation) | UDE · AIST++ | FID_K: 47.21 -> 28.44 | Helped |
| Inject Gaussian noise z into token prediction (ablation) | UDE UTT · HumanML3D | Div_k: 4.28 -> 5.05 | Helped |
| Inject Gaussian noise z into token prediction (ablation) | UDE UTT · HumanML3D | Top-1 Acc.: 8.11 -> 8.33 | Helped |
| Inject Gaussian noise z into token prediction (ablation) | UDE UTT · AIST++ | FID_k: 28.44 -> 27.99 | Helped |
| Replace VQ decoder with diffusion motion decoder (ablation) | UDE · AIST++ | FID_k: 28.44 to 18.43 | Helped |
| Replace VQ decoder with diffusion motion decoder (ablation) | UDE · HumanML3D | Div_k: 4.28 -> 6.98 | Helped |
| Use unified text-and-audio training instead of text-only training (ablation) | UDE · HumanML3D | Top-1 Acc.: 7.77 -> 8.11 | Helped |
| Use unified text-and-audio training instead of text-only training (ablation) | UDE · HumanML3D | FID_k: 31.73 -> 27.66 | Helped |
| Use unified text-and-audio training instead of audio-only training (ablation) | UDE · AIST++ | Beat Align: 0.2231 -> 0.2268 | Helped |
| Train one unified model rather than modality-specific models (ablation) | UDE · AIST++ | FID_m: 11.65 (audio only); 15.70 (unified) | Mixed |
Closest papers by meaning
- Seamless Human Motion Composition with Blended Positional Encodings (CVPR 2024 · similarity 0.89)
- Everything2Motion: Synchronizing Diverse Inputs via a Unified Framework for Human Motion Synthesis (AAAI 2024 · similarity 0.89)
- Rethinking Diffusion for Text-Driven Human Motion Generation: Redundant Representations, Evaluation, and Masked Autoregression (CVPR 2025 · similarity 0.88)
- 🎧MOSPA: Human Motion Generation Driven by Spatial Audio (NeurIPS 2025 · similarity 0.88)
- AMD: Autoregressive Motion Diffusion (AAAI 2024 · similarity 0.88)
- The Language of Motion: Unifying Verbal and Non-verbal Language of 3D Human Motion (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.