Paper 2602.08794
MOVA: Towards Scalable and Synchronized Video-Audio Generation
- Published
- Feb 2026
- Research lab
- Independent
- Citations
- 29
- GitHub
- 1.1K stars
01 In brief
Summary
MOVA is an open-source model for generating synchronized video and audio, including lip-synced speech, sound effects, and music.
It uses a Mixture-of-Experts architecture with 32B total parameters (18B active) and supports Image-Text to Video-Audio generation.
The model couples a pretrained A14B video DiT and a 1.3B audio DiT via a 2.6B bidirectional Bridge module with Aligned RoPE.
Training uses a three-phase curriculum (360p to 720p) with decoupled timestep sampling and heterogeneous learning rates.
A dual classifier-free guidance (CFG) improves alignment.
Evaluations show MOVA outperforms baselines (LTX-2, Ovi, WAN2.1+MMAudio) in audio fidelity, lip-sync, and multi-speaker attribution, achieving an ELO of 1113.8 in human preference.
The data pipeline includes preprocessing, quality filtering, and captioning.
Limitations include degraded performance on singing and complex music, multi-speaker challenges, and high computational cost.
The model weights and code are released to foster research.
02 From the paper
Abstract
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.