MOVA released: 32B MoE end-to-end synchronized video-audio generation model fully open-sourced
Original title:MOVA:可扩展的同步视频-音频生成
AI summary
The OpenMOSS team released MOVA, an audio-video foundation model with 32B total MoE parameters and 18B activated, supporting end-to-end synchronized text- or image-to-audio-video generation at up to 720p and 8 seconds per clip. The model uses an asymmetric dual-tower architecture (a 14B Wan 2.2 I2V video tower and a 1.3B audio tower) plus a bidirectional Bridge and Aligned ROPE for timeline alignment, with model weights, training and inference code, and fine-tuning recipes released.
Key Takeaways
- The OpenMOSS team released MOVA,an audio-video foundation model with 32B total MoE parameters and…
- The model uses an asymmetric dual-tower architecture (a 14B Wan 2.
- 2 I2V video tower and a 1.
- 3B audio tower) plus a bidirectional Bridge and Aligned ROPE for timeline alignment,with model…
Quick Facts
- Published
- Updated
- Category
- Model Releases
- AI score
- 62
原文给出架构、训练策略和开源范围,读者可以了解端到端音视频生成如何处理同步问题并可复现研究。
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: MOVA released: 32B MoE end-to-end synchronized video-audio generation model fully open-sourced. February 1, 2026. https://new.nat6.net/en/news/cmtym5ry70006rou1uyjr6ods/