OpenMOSS has announced the release of MOVA (MOSS-Video-and-Audio), a fully open-source artificial intelligence model designed for video and audio generation. The model stands out for its Mixture of Experts (MoE) architecture, with 18 billion active parameters out of a total of 32 billion.

Technical Details

MOVA offers day-0 support for SGLang-Diffusion, a framework that facilitates the creation of diffusion applications. The pre-trained MOVA-360p and MOVA-720p models are available on Hugging Face.

Resources