Video generation AI “MiniMax H3” debuts, set to be released as an open model with world No. 2-level performance

China’s MiniMax has unveiled its video-generation AI model, MiniMax H3.

It accepts text, images, video, and audio as inputs and can generate up to 15-second videos with sound at 768p or 2K.

Third-party evaluations ranked it highly across multiple video-generation tasks, and it has now been released on Hugging Face and ModelScope after API access.

Open distribution could broaden use cases in filmmaking, advertising, social content, and developer integrations.