Research map / Video / Video generation
TLControl: Trajectory and Language Control for Human Motion Synthesis
Weilin Wan, Zhiyang Dou, Taku Komura, Wenping Wang, Dinesh Jayaraman, Lingjie Liu
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 |
|---|---|---|---|
| use part-based VQ-VAE instead of unsplit full-body VQ-VAE (ablation) | TLControl · HumanML3D test (all joints control) | FID: 0.066 -> 0.032 | Improved |
| use part-based VQ-VAE instead of unsplit VQ-VAE (ablation) | TLControl · HumanML3D test | runtime: 27.3% decrease | Improved |
| optimize in latent space instead of joint-level IK (ablation) | TLControl · HumanML3D root control, 50% masked | FID: 0.675 -> 0.311 | Improved |
| add latent test-time optimization to MTT output (ablation) | TLControl · HumanML3D root control, 0% mask | FID: 0.343 -> 0.271 | Improved |
| apply joint-level IK on top of MTT output for left hand control (ablation) | TLControl · HumanML3D left hand control, 0% mask | FID: 0.226 -> 0.383 | Hurt |
| tighten optimization accuracy criterion from 1E-4 to 1E-8 (ablation) | TLControl · HumanML3D all joints | Average error (cm): 6.72 -> 0.54 (runtime 3.48 s -> 127.90 s per batch) | Improved |
| increase trajectory masking rate for root trajectory (ablation) | TLControl · HumanML3D | MModality: 0.994 -> 1.982 | Improved |
| replace multi-step diffusion sampling with feed-forward MTT plus latent optimization | TLControl vs OmniControl · HumanML3D | time (s/frame): 0.606 -> 0.015 | Improved |
| TLControl versus OmniControl on foot-joint FID | TLControl · HumanML3D left foot control | FID: 0.280 -> 0.368 | Hurt |
Closest papers by meaning
- MotionCtrl: A Real-time Controllable Vision-Language-Motion Model (ICCV 2025 · similarity 0.88)
- DartControl: A Diffusion-Based Autoregressive Motion Model for Real-Time Text-Driven Motion Control (ICLR 2025 · similarity 0.87)
- InterControl: Zero-shot Human Interaction Generation by Controlling Every Joint (NeurIPS 2024 · similarity 0.87)
- Enhanced Fine-Grained Motion Diffusion for Text-Driven Human Motion Synthesis (AAAI 2024 · similarity 0.87)
- WANDR: Intention-guided Human Motion Generation (CVPR 2024 · similarity 0.86)
- NIFTY: Neural Object Interaction Fields for Guided Human Motion Synthesis (CVPR 2024 · similarity 0.86)
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.
