Research map / Video / Video generation
UniAVGen: Unified Audio and Video Generation with Asymmetric Cross-Modal Interactions
Guozhen Zhang, Zixiang Zhou, Teng Hu, Ziqiao Peng, Youliang Zhang, Yi Chen, Yuan Zhou, Qinglin Lu, Limin Wang
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 temporal-aligned symmetric interaction instead of global interaction (ablation) | UniAVGen · Stage 2 first 10k steps | LS: 3.46 -> 3.73 | Improved |
| use asymmetric temporal-aligned interaction in both directions instead of symmetric (STI) (ablation) | UniAVGen · Stage 2 first 10k steps | LS: 3.73 -> 4.09 | Improved |
| use asymmetric interaction in V2A only (ablation) | UniAVGen · Stage 2 first 10k steps | LS: 3.73 -> 3.88 | Improved |
| use asymmetric interaction in A2V only (ablation) | UniAVGen · Stage 2 first 10k steps | LS: 3.73 -> 3.97 | Improved |
| add supervised face-aware modulation (ablation) | UniAVGen · Stage 2 first 10k steps | LS: 3.89 -> 4.09 | Improved |
| add face-aware modulation without mask supervision (ablation) | UniAVGen · Stage 2 first 10k steps | LS: 3.89 -> 3.92 | No clear effect |
| decay mask loss weight over training instead of fixing it (ablation) | UniAVGen FAM · Stage 2 first 10k steps | EC: 0.497 -> 0.504 | Improved |
| decay mask loss weight instead of fixed (ablation) | UniAVGen FAM · Stage 2 first 10k steps | LS: 4.11 -> 4.09 | No clear effect |
| add modality-aware classifier-free guidance at inference (ablation) | UniAVGen · qualitative joint generation | emotion and motion alignment | Improved |
| train joint generation first, then multi-task learning (versus joint generation only) (ablation) | UniAVGen · training strategy comparison | LC | Improved |
| multi-task training from the start (versus joint generation first) (ablation) | UniAVGen · training strategy comparison | convergence speed | Hurt |
Later papers that compared with this method
1 later accepted paper reports UniAVGen as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- JoVA: Unified Multimodal Learning for Joint Video-Audio Generation and Editing (ECCV 2026 · better on 3 of 4 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
- UniFlow-Audio: Unified Flow Matching for Audio Generation from Omni-Modalities (NeurIPS 2026 · similarity 0.90)
- MMDisCo: Multi-Modal Discriminator-Guided Cooperative Diffusion for Joint Audio and Video Generation (ICLR 2025 · 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 0.88)
- Model-Guided Dual-Role Alignment for High-Fidelity Open-Domain Video-to-Audio Generation (NeurIPS 2025 · similarity 0.88)
- Seeing and Hearing: Open-domain Visual-Audio Generation with Diffusion Latent Aligners (CVPR 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.
