Research map / Video / Video generation
Long-Term Rhythmic Video Soundtracker
Jiashuo Yu, Yaohui Wang, Xinyuan Chen, Xiao Sun, Yu Qiao
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 context-aware latent diffusion with visual and rhythm conditioning | LORIS · LORIS_FE25 | F1: 54.3 -> 62.7 | Improved |
| Add visual conditioning (ablation) | LORIS · LORIS_FE25 | F1: 49.6 -> 62.7 | Improved |
| Add rhythm conditioning (ablation) | LORIS · LORIS_FE25 | F1: 45.8 -> 62.7 | Improved |
| Use RGB-only visual features instead of additionally concatenating ST-GCN motion features (ablation) | LORIS · LORIS_FE25 | F1: 61.0 -> 62.7 | Improved |
| Add Bi-LSTM temporal modeling of visual features (ablation) | LORIS · LORIS_FE25 | F1: 60.0 -> 62.7 | Improved |
| Add Hawkes process to rhythm positional encoding (ablation) | LORIS · LORIS_FE25 | F1: 54.9 -> 62.7 | Improved |
| Add positional encoding to rhythm embeddings (ablation) | LORIS · LORIS_FE25 | F1: 57.6 -> 62.7 | Improved |
| Embed and cross-attend rhythm peaks instead of multiplying rhythm envelopes with audio latents (ablation) | LORIS · LORIS_FE25 | F1: 58.7 -> 62.7 | Improved |
| Increase inference sampling steps from 10 to 50 (ablation) | LORIS · LORIS_FE25 | F1: 55.8 -> 62.7 | Improved |
| Increase inference sampling steps from 50 to 100 (ablation) | LORIS · LORIS_FE25 | F1: 62.7 -> 62.9 | No clear effect |
| Set classifier-free guidance scale to 20 rather than tested alternatives (ablation) | LORIS · LORIS_FE25 | F1: 56.6, 59.4, 57.2, 62.6, 61.2, 60.2; w=20: 62.7 | Improved |
Closest papers by meaning
- MoMu-Diffusion: On Learning Long-Term Motion-Music Synchronization and Correspondence (NeurIPS 2024 · similarity 0.90)
- Customized Condition Controllable Generation for Video Soundtrack (CVPR 2025 · similarity 0.89)
- DanceAnyWay: Synthesizing Beat-Guided 3D Dances with Randomized Temporal Contrastive Learning (AAAI 2024 · similarity 0.89)
- V2Meow: Meowing to the Visual Beat via Video-to-Music Generation (AAAI 2024 · similarity 0.88)
- VidMuse: A Simple Video-to-Music Generation Framework with Long-Short-Term Modeling (CVPR 2025 · similarity 0.87)
- Bidirectional Autoregessive Diffusion Model for Dance 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.
