Research map / Video / Video generation
Music-Aligned Holistic 3D Dance Generation via Hierarchical Motion Modeling
Xiaojie Li, Ronghui Li, Shukai Fang, Shuzhao Xie, Xiaoyang Guo, Jiaqing Zhou, Junkun Peng, Zhi Wang
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 hierarchical residual vector quantization instead of RVQ-5 | motion tokenizer · SoulDance motion reconstruction (MPJPE all, mm) | MPJPE all (mm): 108.983 -> 83.679 | Improved |
| use RVQ instead of plain VQ | motion tokenizer · SoulDance motion reconstruction (MPJPE all, mm) | MPJPE all (mm): 136.752 -> 108.983 | Improved |
| use HRVQ instead of RVQ for generation (ablation) | SoulNet w/o MMR · SoulDance generation ablation (HRVQ/MMR) | FID: 0.082 -> 0.048 | Improved |
| add MMR supervision (ablation) | SoulNet · SoulDance generation ablation (HRVQ/MMR) | FID: 0.048 -> 0.029 | Improved |
| add MMR supervision (ablation) | SoulNet · SoulDance (dance generation) | MMR-MS: 0.418 -> 0.369 | Improved |
| add MMR supervision (ablation) | SoulNet · SoulDance (dance generation) | EAS: 0.435 -> 0.594 | Improved |
| add MMR supervision to RVQ (ablation) | RVQ-512 · SoulDance generation ablation (HRVQ/MMR) | FID: 0.082 -> 0.067 | Improved |
| increase RVQ codebook size from 512 to 1024 (ablation) | RVQ + MMR · SoulDance generation ablation (HRVQ/MMR) | FID: 0.067 -> 0.090 | Hurt |
| add L_Align-whole to L_Align-body (ablation) | SoulNet · SoulDance MMR loss ablation | MMR-MS: 0.387 -> 0.369 | No clear effect |
| add L_Align-body to L_Align-whole (ablation) | SoulNet · SoulDance MMR loss ablation | FID: 0.042 -> 0.029 | Improved |
| increase HRVQ residual layers from V=1 to V=6 (ablation) | HRVQ · HRVQ layers V (SoulDance reconstruction, MPJPE all) | MPJPE all (mm): 104.859 -> 79.518 | Improved |
| increase HRVQ residual layers from V=6 to V=7 (ablation) | HRVQ · HRVQ layers V (SoulDance reconstruction, MPJPE all) | MPJPE all (mm): 79.518 -> 84.080 | Hurt |
Closest papers by meaning
- OpenDance: Multimodal Controllable 3D Dance Generation with Large-scale Internet Data (CVPR 2026 · similarity 0.92)
- OmniDance: Multimodal Driven Dance Video Generation with Large-scale Internet Data (ECCV 2026 · similarity 0.91)
- DanceAnyWay: Synthesizing Beat-Guided 3D Dances with Randomized Temporal Contrastive Learning (AAAI 2024 · similarity 0.91)
- MEGADance: Mixture-of-Experts Architecture for Genre-Aware 3D Dance Generation (NeurIPS 2025 · similarity 0.90)
- DanceCamera3D: 3D Camera Movement Synthesis with Music and Dance (CVPR 2024 · similarity 0.89)
- Bidirectional Autoregessive Diffusion Model for Dance Generation (CVPR 2024 · similarity 0.89)
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.
