Research map / Video / Video generation
Multi-Modal Controlled Coherent Motion Generation
Yifei Liu, Qiong Cao, Hongwei Yi, Huaiguang Jiang, Changxing Ding
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 |
|---|---|---|---|
| add spine to the audio-driven body mask (ablation) | MOCO · Multi-modal benchmark | FID+: 0.862 -> 0.921 | Hurt |
| use spine and legs for the audio-driven mask instead of head and arms (ablation) | MOCO · Multi-modal benchmark | R1: 24.6 -> 7.9 | Hurt |
| use spine and legs for audio mask instead of head and arms (ablation) | MOCO · Multi-modal benchmark | FID-A: 3.83 -> 2.75 | Mixed |
| share weights between text and speech denoisers (ablation) | MOCO · Multi-modal benchmark | R1: 24.6 -> 22.1 | Hurt |
| share weights between text and speech denoisers (ablation) | MOCO · Multi-modal benchmark | FID-A: 3.83 -> 4.24 | Hurt |
| add per-step combination of body parts instead of combining only at the end (ablation) | MOCO · Multi-modal benchmark | TSR: 1.17 -> 0.90 | Improved |
| add per-step combination instead of combining only at the end (ablation) | MOCO · Multi-modal benchmark | Beat consistency: 2.27 -> 2.72 | Improved |
| add per-step combination instead of combining once (ablation) | MOCO · User study | Better body coherence (%): 17.0 -> 71.0 | Improved |
| use decoupled denoising instead of weighted sum | MOCO vs Weighted Sum · Multi-modal benchmark | R1: 6.8 -> 24.6 | Improved |
| use decoupled denoising instead of pseudo-text training | MOCO vs Pseudo-Text · Multi-modal benchmark | R1: 2.2 -> 24.6 | Improved |
| add post-hoc L-BFGS optimization to trajectory predictor (ablation) | G_T2V · Trajectory control | Location average difference: 0.56 -> 0.08 | Improved |
| add classifier-free guidance to trajectory predictor (ablation) | G_T2V · Trajectory control | Location average difference: 0.56 -> 0.57 | No clear effect |
Closest papers by meaning
- 🎧MOSPA: Human Motion Generation Driven by Spatial Audio (NeurIPS 2025 · similarity 0.90)
- Seamless Human Motion Composition with Blended Positional Encodings (CVPR 2024 · similarity 0.88)
- AMD: Autoregressive Motion Diffusion (AAAI 2024 · similarity 0.88)
- MotionCraft: Crafting Whole-Body Motion with Plug-and-Play Multimodal Controls (AAAI 2025 · similarity 0.88)
- DAWN: Dynamic Frame Avatar with Non-autoregressive Diffusion Framework for Talking head Video Generation (ICLR 2025 · similarity 0.87)
- Towards Variable and Coordinated Holistic Co-Speech Motion Generation (CVPR 2024 · 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.
