Research map / Video / Video generation
Precise Video-to-Audio Generation with Cross-Modal Alignment in Latent Space
Thanh V. T. Tran, Ngoc-Son Nguyen, Luong Tran, Long-Khanh Pham, Paarth Neekhara, Shehzeen Hussain, Van Nguyen
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 Progressive Soft-masked Cross-Attention (replace standard visual cross-attention) (ablation) | Flowley · VGGSound test | Align Acc: 87.61 -> 89.37 | Improved |
| add layer-dependent progressive parameter beta (ablation) | Flowley PSCA · VGGSound test | Align Acc: 88.62 -> 89.37 | Improved |
| add layer-dependent progressive parameter beta (ablation) | Flowley PSCA · VGGSound test | IS: 17.78 -> 18.25 | Improved |
| add text and visual cross-attention to single-stream block (vs vanilla DiT) (ablation) | Flowley · VGGSound test | IS: 16.99 -> 18.25 | Improved |
| add text cross-attention (ablation) | Flowley · VGGSound test | IB-Score: 28.51 -> 29.32 | Improved |
| widen PSCA window omega and delta (ablation) | Flowley · VGGSound test | Align Acc: 89.37 -> 86.58 | Hurt |
| widen PSCA window omega and delta (ablation) | Flowley · VGGSound test | IS / IB-Score: IS 17.85-18.57, IB 28.84-29.48 | No clear effect |
| add SoundCap captions | Flowley · VGGSound test | IS: 18.25 -> 19.68 | Improved |
| add SoundCap captions | MMAudio · VGGSound test | KAD: 0.57 -> 0.39 | Improved |
| add SoundCap captions without noise-aware prompts (ablation) | Flowley · VGGSound test | IS: 18.25 -> 15.16 | Hurt |
| add noise-aware conditions to SoundCap prompts (ablation) | Flowley + SoundCap · VGGSound test | IS: 15.16 -> 19.68 | Improved |
| add SoundCap at inference | Flowley · MovieGen Audio Bench (zero-shot) | IS: 7.74 -> 8.18 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- VinTAGe — VinTAGe: Joint Video and Text Conditioning for Holistic Audio Generation (CVPR 2025) · better on 3 of 3 shared results
- Frieren — Frieren: Efficient Video-to-Audio Generation Network with Rectified Flow Matching (NeurIPS 2024) · better on 0 of 1 shared result
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
- Tell What You Hear From What You See - Video to Audio Generation Through Text (NeurIPS 2024 · similarity 0.90)
- MMAudio: Taming Multimodal Joint Training for High-Quality Video-to-Audio Synthesis (CVPR 2025 · similarity 0.90)
- From Vision to Audio and Beyond: A Unified Model for Audio-Visual Representation and Generation (ICML 2024 · similarity 0.89)
- Native Audio-Visual Alignment for Generation (NeurIPS 2026 · similarity 0.89)
- VinTAGe: Joint Video and Text Conditioning for Holistic Audio Generation (CVPR 2025 · similarity 0.89)
- Seeing and Hearing: Open-domain Visual-Audio Generation with Diffusion Latent Aligners (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.
