Research map / Video / Video generation
Gotta Hear Them All: Towards Sound Source Aware Audio Generation
Wei Guo, Heng Wang, Jianbo Ma, Weidong Cai
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 CMSS manifold embeddings to the Remixer input (ablation) | SS2A · VGG-SS multi-source | SSMS: 1.579 -> 5.973 | Improved |
| add CMSS manifold embeddings to the Remixer input (ablation) | SS2A · VGG-SS single-source | C-FAD: 122.127 -> 15.150 | Improved |
| add CLIP embeddings alongside CMSS in the Remixer input (ablation) | SS2A · VGG-SS single-source | C-FAD: 96.045 -> 15.150 | Improved |
| add CLIP embeddings alongside CMSS in the Remixer input (ablation) | SS2A · VGG-SS multi-source | SSMS: 5.553 -> 5.973 | Improved |
| add CCMR mask regularization (alpha=0.35) (ablation) | SS2A · VGG-SS multi-source | SSMS: 1.342 -> 5.973 | Improved |
| increase CCMR alpha from 0.35 to 1 (ablation) | SS2A · VGG-SS multi-source | SSMS: 5.973 -> 4.237 | Hurt |
| add Cycle Mix (sample size 64, iterations 64) (ablation) | SS2A · VGG-SS multi-source | SSMS: 5.868 -> 5.973 | Improved |
| add Cycle Mix (sample size 64, iterations 64) (ablation) | SS2A · VGG-SS single-source | SSMS: 5.048 -> 4.936 | Hurt |
| use 4 attention layers instead of 1 in the Remixer (ablation) | SS2A · VGG-SS multi-source | SSMS: 5.973 -> 5.395 | Hurt |
| use 4 attention layers instead of 1 in the Remixer (ablation) | SS2A · VGG-SS single-source | C-FAD: 15.150 -> 14.021 | Improved |
| use CMSS variant B instead of shallower variant A (ablation) | SS2A · VGG-SS multi-source | SSMS: 4.868 -> 5.973 | Improved |
| use CMSS variant B instead of deeper variant C (ablation) | SS2A · VGG-SS multi-source | SSMS: 4.763 -> 5.973 | Improved |
Closest papers by meaning
- V2A-Mapper: A Lightweight Solution for Vision-to-Audio Generation by Connecting Foundation Models (AAAI 2024 · similarity 0.90)
- SoundVista: Novel-View Ambient Sound Synthesis via Visual-Acoustic Binding (CVPR 2025 · similarity 0.89)
- Sounding that Object: Interactive Object-Aware Image to Audio Generation (ICML 2025 · similarity 0.89)
- Read, Watch and Scream! Sound Generation from Text and Video (AAAI 2025 · similarity 0.88)
- Object-aware Sound Source Localization via Audio-Visual Scene Understanding (CVPR 2025 · similarity 0.88)
- Audible World Models: Spatially Aware Sound Generation for 3D Worlds (NeurIPS 2026 · 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.
