Research map / Video / Video generation
JavisDiT: Joint Audio-Video Diffusion Transformer with Hierarchical Spatio-Temporal Prior Synchronization
Kai Liu, Wei Li, Lai Chen, Shengqiong Wu, Yanhao Zheng, Jiayi Ji, Fan Zhou, Jiebo Luo, Ziwei Liu, Hao Fei, Tat-Seng Chua
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 |
|---|---|---|---|
| replace UNet audio backbone with STDiT (ablation) | JavisDiT · JavisBench-mini (ablation) | normalized AV-Quality (lower better): 9.371 -> 7.293 | Improved |
| add HiST-Sypo estimator (ablation) | JavisDiT · JavisBench-mini (ablation) | normalized AV-Synchrony: 0.130 -> 0.150 | Improved |
| add HiST-Sypo estimator (ablation) | JavisDiT · JavisBench-mini (ablation) | normalized AV-Quality (lower better): 7.293 -> 6.127 | Improved |
| add bidirectional cross-attention (BiCA) (ablation) | JavisDiT · JavisBench-mini (ablation) | normalized AV-Synchrony: 0.130 -> 0.133 | No clear effect |
| add BiCA on top of HiST-Sypo (ablation) | JavisDiT · JavisBench-mini (ablation) | normalized AV-Synchrony: 0.150 -> 0.153 | No clear effect |
| increase ST-prior tokens from 1 to 32 (ablation) | JavisDiT · JavisBench-mini (ablation) | normalized AV-Synchrony: 0.137 -> 0.153 | Improved |
| inject ST-priors via cross-attention instead of addition (ablation) | JavisDiT · JavisBench-mini (ablation) | normalized AV-Synchrony: 0.144 -> 0.153 | Improved |
| inject ST-priors via cross-attention instead of modulation (ablation) | JavisDiT · JavisBench-mini (ablation) | normalized AV-Synchrony: 0.145 -> 0.153 | Improved |
| use 1 prior token (ablation) | JavisDiT · JavisBench-mini (ablation) | normalized AV-Quality (lower better): 6.581 -> 6.909 | Hurt |
| joint end-to-end JAVG instead of cascaded T2V+V2A | JavisDiT · JavisBench (4s, 240P) | JavisScore: 0.151 -> 0.154 | Improved |
| use OpenSora backbone instead of Wan2.1 | JavisDiT vs UniVerse-1 · JavisBench (4s, 240P) | FVD / video-text alignment: 204.1 vs 194.2 FVD | Hurt |
| generate 10s instead of 4s videos | JavisDiT · JavisBench-mini | JavisScore: 0.153 -> 0.154 | No clear effect |
Later papers that compared with this method
5 later accepted papers report JavisDiT as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- Harmony: Harmonizing Audio and Video Generation through Cross-Task Synergy (CVPR 2026 · better on 8 of 8 shared results)
- OmniForcing: Unleashing Real-time Joint Audio-Visual Generation (ECCV 2026 · better on 3 of 4 shared results)
- JavisDiT++: Unified Modeling and Optimization for Joint Audio-Video Generation (ICLR 2026 · better on 3 of 3 shared results)
- TMD-Bench: A Multi-Level Evaluation Paradigm for Music–Dance Co-Generation (ICML 2026 · better on 2 of 2 shared results)
- JoVA: Unified Multimodal Learning for Joint Video-Audio Generation and Editing (ECCV 2026 · 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
- Seeing and Hearing: Open-domain Visual-Audio Generation with Diffusion Latent Aligners (CVPR 2024 · similarity 0.88)
- MMAudio: Taming Multimodal Joint Training for High-Quality Video-to-Audio Synthesis (CVPR 2025 · similarity 0.87)
- MMDisCo: Multi-Modal Discriminator-Guided Cooperative Diffusion for Joint Audio and Video Generation (ICLR 2025 · similarity 0.87)
- Diverse and Aligned Audio-to-Video Generation via Text-to-Video Model Adaptation (AAAI 2024 · similarity 0.86)
- Native Audio-Visual Alignment for Generation (NeurIPS 2026 · similarity 0.86)
- A Versatile Diffusion Transformer with Mixture of Noise Levels for Audiovisual Generation (NeurIPS 2024 · similarity 0.86)
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.
