Research map / Video / Video generation
X-Dancer: Expressive Music to Human Dance Video Generation
Zeyuan Chen, Hongyi Xu, Guoxian Song, You Xie, Chenxu Zhang, Xin Chen, Chao Wang, Di Chang, Linjie Luo
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 global music context (ablation) | X-Dancer transformer · In-House motion | FVD: 265.73 -> 238.22 | Improved |
| add global motion context (ablation) | X-Dancer transformer · In-House motion | FVD: 247.54 -> 238.22 | Improved |
| add global music context (ablation) | X-Dancer transformer · In-House motion | BAS: 0.2142 -> 0.2182 | Improved |
| scale data from 10k to full 76k videos and model from GPT-medium (ablation) | X-Dancer transformer · In-House motion | FVD: 402.63 -> 238.22 | Improved |
| scale model from GPT-small (117M) to GPT-medium (345M) on 10k subset (ablation) | X-Dancer transformer · In-House motion | FVD: 332.93 -> 402.63 | Hurt |
| use multi-part instead of single-part VQ-VAE tokenization (ablation) | X-Dancer pose tokenizer · Pose reconstruction | Pose L1 full body: 0.83 -> 0.50 | Improved |
| use multi-part instead of single-part VQ-VAE tokenization (ablation) | X-Dancer video synthesis · Self-driven video | PSNR: 17.079 -> 19.148 | Improved |
| use motion token decoder instead of pose guider (ablation) | X-Dancer video synthesis · In-House video synthesis | FVD: 735.05 -> 507.06 | Improved |
| use motion token decoder instead of pose guider (ablation) | X-Dancer video synthesis · In-House video synthesis | ID-SIM: 0.4894 -> 0.5317 | Improved |
| use motion token decoder instead of pose guider (ablation) | Multi-part VQ-VAE · Self-driven video | FVD: 384.16 -> 295.87 | Improved |
| train motion transformer on AIST instead of in-house data (ablation) | X-Dancer · In-House video synthesis | FVD: 507.06 -> 549.38 | Hurt |
Closest papers by meaning
- DanceCamera3D: 3D Camera Movement Synthesis with Music and Dance (CVPR 2024 · similarity 0.90)
- DanceAnyWay: Synthesizing Beat-Guided 3D Dances with Randomized Temporal Contrastive Learning (AAAI 2024 · similarity 0.90)
- Bidirectional Autoregessive Diffusion Model for Dance Generation (CVPR 2024 · similarity 0.88)
- Breaking the Group Size Barrier: Parameter-Efficient Group Dance Generation with Chain-of-Dancers (NeurIPS 2026 · similarity 0.88)
- Controllable 3D Dance Generation Using Diffusion-Based Transformer U-Net (AAAI 2025 · similarity 0.88)
- DisCo: Disentangled Control for Realistic Human Dance Generation (CVPR 2024 · similarity 0.88)
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.
