MOVA: Towards Scalable and Synchronized Video-Audio Generation
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866908824879759360 |
|---|---|
| author | OpenMOSS Team Yu, Donghua Chen, Mingshu Chen, Qi Luo, Qi Wu, Qianyi Cheng, Qinyuan Li, Ruixiao Liang, Tianyi Zhang, Wenbo Tu, Wenming Peng, Xiangyu Gao, Yang Huo, Yanru Zhu, Ying Luo, Yinze Zhang, Yiyang Song, Yuerong Xu, Zhe Zhang, Zhiyu Yang, Chenchen Chang, Cheng Zhou, Chushu Chen, Hanfu Ma, Hongnan Li, Jiaxi Tong, Jingqi Liu, Junxi Chen, Ke Li, Shimin Jiang, Shiqi Wang, Songlin Jiang, Wei Fei, Zhaoye Ning, Zhiyuan Li, Chunguo Li, Chenhui He, Ziwei Huang, Zengfeng Chen, Xie Qiu, Xipeng |
| author_facet | OpenMOSS Team Yu, Donghua Chen, Mingshu Chen, Qi Luo, Qi Wu, Qianyi Cheng, Qinyuan Li, Ruixiao Liang, Tianyi Zhang, Wenbo Tu, Wenming Peng, Xiangyu Gao, Yang Huo, Yanru Zhu, Ying Luo, Yinze Zhang, Yiyang Song, Yuerong Xu, Zhe Zhang, Zhiyu Yang, Chenchen Chang, Cheng Zhou, Chushu Chen, Hanfu Ma, Hongnan Li, Jiaxi Tong, Jingqi Liu, Junxi Chen, Ke Li, Shimin Jiang, Shiqi Wang, Songlin Jiang, Wei Fei, Zhaoye Ning, Zhiyuan Li, Chunguo Li, Chenhui He, Ziwei Huang, Zengfeng Chen, Xie Qiu, Xipeng |
| contents | Audio is indispensable for real-world video, yet generation models have largely overlooked audio components. Current approaches to producing audio-visual content often rely on cascaded pipelines, which increase cost, accumulate errors, and degrade overall quality. While systems such as Veo 3 and Sora 2 emphasize the value of simultaneous generation, joint multimodal modeling introduces unique challenges in architecture, data, and training. Moreover, the closed-source nature of existing systems limits progress in the field. In this work, we introduce MOVA (MOSS Video and Audio), an open-source model capable of generating high-quality, synchronized audio-visual content, including realistic lip-synced speech, environment-aware sound effects, and content-aligned music. MOVA employs a Mixture-of-Experts (MoE) architecture, with a total of 32B parameters, of which 18B are active during inference. It supports IT2VA (Image-Text to Video-Audio) generation task. By releasing the model weights and code, we aim to advance research and foster a vibrant community of creators. The released codebase features comprehensive support for efficient inference, LoRA fine-tuning, and prompt enhancement. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_08794 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | MOVA: Towards Scalable and Synchronized Video-Audio Generation OpenMOSS Team Yu, Donghua Chen, Mingshu Chen, Qi Luo, Qi Wu, Qianyi Cheng, Qinyuan Li, Ruixiao Liang, Tianyi Zhang, Wenbo Tu, Wenming Peng, Xiangyu Gao, Yang Huo, Yanru Zhu, Ying Luo, Yinze Zhang, Yiyang Song, Yuerong Xu, Zhe Zhang, Zhiyu Yang, Chenchen Chang, Cheng Zhou, Chushu Chen, Hanfu Ma, Hongnan Li, Jiaxi Tong, Jingqi Liu, Junxi Chen, Ke Li, Shimin Jiang, Shiqi Wang, Songlin Jiang, Wei Fei, Zhaoye Ning, Zhiyuan Li, Chunguo Li, Chenhui He, Ziwei Huang, Zengfeng Chen, Xie Qiu, Xipeng Computer Vision and Pattern Recognition Sound Audio is indispensable for real-world video, yet generation models have largely overlooked audio components. Current approaches to producing audio-visual content often rely on cascaded pipelines, which increase cost, accumulate errors, and degrade overall quality. While systems such as Veo 3 and Sora 2 emphasize the value of simultaneous generation, joint multimodal modeling introduces unique challenges in architecture, data, and training. Moreover, the closed-source nature of existing systems limits progress in the field. In this work, we introduce MOVA (MOSS Video and Audio), an open-source model capable of generating high-quality, synchronized audio-visual content, including realistic lip-synced speech, environment-aware sound effects, and content-aligned music. MOVA employs a Mixture-of-Experts (MoE) architecture, with a total of 32B parameters, of which 18B are active during inference. It supports IT2VA (Image-Text to Video-Audio) generation task. By releasing the model weights and code, we aim to advance research and foster a vibrant community of creators. The released codebase features comprehensive support for efficient inference, LoRA fine-tuning, and prompt enhancement. |
| title | MOVA: Towards Scalable and Synchronized Video-Audio Generation |
| topic | Computer Vision and Pattern Recognition Sound |
| url | https://arxiv.org/abs/2602.08794 |