Research map / Video / Video generation
Omni2Sound: Towards Unified Video-Text-to-Audio Generation
Yusheng Dai, Zehua Chen, Yuxuan Jiang, Qiuhong Ke, Jianfei Cai, Jun Zhu
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 |
|---|---|---|---|
| raise T2A sampling ratio from 0.20 to 0.40 without VT2A data (ablation) | Omni2Sound Stage 2 · VGGSound-Omni T2A | FAD: 1.36 -> 1.06 | Improved |
| raise T2A sampling ratio from 0.20 to 0.40 without VT2A data (ablation) | Omni2Sound Stage 2 · VGGSound-Omni V2A | FAD: 0.56 -> 0.62 | Hurt |
| add high-alignment SoundAtlas VT2A data as bridge (ablation) | Omni2Sound Stage 2 · VGGSound-Omni T2A | FAD: 1.06 -> 0.94 | Improved |
| use SoundAtlas captions instead of audio-only Gemini 2.5 captions (ablation) | Omni2Sound Stage 2 · VGGSound-Omni T2A | FAD: 1.13 -> 0.94 | Improved |
| add Stage 1 T2A pretraining (ablation) | Omni2Sound · VGGSound-Omni T2A | FAD: 1.22 -> 0.94 | Improved |
| add Stage 1 T2A pretraining (ablation) | Omni2Sound · VGGSound-Omni V2A | FAD: 0.68 -> 0.57 | Improved |
| add decoupled Stage 3 robustness training after Stage 2 (ablation) | Omni2Sound · VGGSound-Omni V2A | FAD: 0.57 -> 0.51 | Improved |
| add decoupled Stage 3 robustness training after Stage 2 (ablation) | Omni2Sound · VGGSound-Omni V2A | DS: 0.49 -> 0.47 | Improved |
| merge Stage 3 into Stage 2 instead of decoupling (ablation) | Omni2Sound · VGGSound-Omni V2A | FAD: 0.51 -> 0.60 | Hurt |
| add Stage 3 off-screen synthesis (ablation) | Omni2Sound · VGGSound-Omni off-screen VT2A | FAD: 0.97 -> 0.85 | Improved |
| add Stage 3 off-screen synthesis (ablation) | Omni2Sound · VGGSound-Omni off-screen VT2A | LA-CLAP: 0.31 -> 0.32 | Improved |
| use Video-LLaMA captions instead of SoundAtlas-style captions at test time | Omni2Sound · VGGSound-Omni VT2A | FAD: 0.53 -> 0.66 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- AudioX — AudioX: A Unified Framework for Anything-to-Audio Generation (ICLR 2026) · better on 5 of 5 shared results
- Frieren — Frieren: Efficient Video-to-Audio Generation Network with Rectified Flow Matching (NeurIPS 2024) · better on 1 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
- VinTAGe: Joint Video and Text Conditioning for Holistic Audio Generation (CVPR 2025 · similarity 0.91)
- Tell What You Hear From What You See - Video to Audio Generation Through Text (NeurIPS 2024 · similarity 0.89)
- AVGen-Bench: A Task-Driven Benchmark for Multi-Granular Evaluation of Text-to-Audio-Video Generation (ICML 2026 · similarity 0.89)
- Diverse and Aligned Audio-to-Video Generation via Text-to-Video Model Adaptation (AAAI 2024 · similarity 0.89)
- V2A-Mapper: A Lightweight Solution for Vision-to-Audio Generation by Connecting Foundation Models (AAAI 2024 · similarity 0.89)
- MMAudio: Taming Multimodal Joint Training for High-Quality Video-to-Audio Synthesis (CVPR 2025 · 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.
