Research map / Video / Video generation
Text Dictates, Music Decorates: Energy-based Attention for Editable Dance Generation
Seong Jong Yoo, Siyuan Peng, Felix Gu, Stratis Aloimonos, Cornelia Fermuller
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 music alignment energy and drift (gamma_m=1, gamma_d=1 vs 0,0) (ablation) | STREAM · Motorica++ | BAS: 0.2301 -> 0.2528 | Improved |
| add Text-AdaLN (ablation) | STREAM · Motorica++ | BAS: 0.2285 -> 0.2528 | Improved |
| add normalization of the music alignment drift (ablation) | STREAM · Motorica++ | FID_k: 10.37 -> 7.32; BAS 0.2251 -> 0.2528 | Improved |
| add MAP update of the music condition (ablation) | STREAM · Motorica++ | BAS: 0.2528 -> 0.2372 | Hurt |
| set music drift weight gamma_d to 1.0 from 0.0 (gamma_m=1) (ablation) | STREAM · Motorica++ | BAS: 0.2369 -> 0.2528 | Improved |
| enable gamma_m=1.0 and gamma_d=1.0 versus gamma_m=0.0, gamma_d=0.5 (ablation) | STREAM · Motorica++ | BAS: 0.2468 -> 0.2528 | Improved |
| use text Bayesian update rate 0.002 instead of none (ablation) | STREAM · Motorica++ | FID_g: 10.73 -> 6.87; BAS 0.2444 -> 0.2528 | Improved |
| increase text Bayesian update rate to 0.1 from 0.002 (ablation) | STREAM · Motorica++ | FID_k: 7.32 -> 50.40 | Hurt |
| add music condition to text-only STREAM | STREAM · Motorica++ (Exchange protocol) | EDS: 0.6207 -> 0.6539 | Improved |
| add text condition to audio-only STREAM | STREAM · Motorica++ (Exchange protocol) | EDS: 0.5853 -> 0.6539 | Improved |
| use BEAM instead of standard cross-attention with AdaLN (ablation) | STREAM · Motorica++ (Exchange protocol) | EDS: 0.5641 -> 0.6539; FID_k 236.88 -> 7.32 | Improved |
| use BEAM instead of late fusion (ablation) | STREAM · Motorica++ (Exchange protocol) | EDS: 0.5887 -> 0.6539; FID_k 20.22 -> 7.32 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Lodge — Lodge++: High-Quality and Long Dance Generation With Robust Choreography Patterns (TPAMI 2025) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- TMD-Bench: A Multi-Level Evaluation Paradigm for Music–Dance Co-Generation (ICML 2026 · similarity 0.92)
- Music-to-Dance Generation via Atomic Movements (ECCV 2026 · similarity 0.92)
- MEGADance: Mixture-of-Experts Architecture for Genre-Aware 3D Dance Generation (NeurIPS 2025 · similarity 0.90)
- Bidirectional Autoregessive Diffusion Model for Dance Generation (CVPR 2024 · similarity 0.89)
- DanceAnyWay: Synthesizing Beat-Guided 3D Dances with Randomized Temporal Contrastive Learning (AAAI 2024 · similarity 0.88)
- MoMu-Diffusion: On Learning Long-Term Motion-Music Synchronization and Correspondence (NeurIPS 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.
