MiniMax Unveils Open-Weight H3 Video AI Model, Challenging ByteDance
The H3 video generation model has been introduced by the Chinese AI company MiniMax, positioning itself against ByteDance's Seedance 2.0. According to Artificial Analysis, H3 is touted as the most advanced video editing tool globally, aimed at sectors such as film, advertising, video games, and product design. Users can access it through an API and the MiniMax Hailuo AI platform. To foster customization across different AI hardware, MiniMax plans to open-source H3's weights, offering them under the MiniMax Community License for free non-commercial use and for organizations with revenues below US$20 million. Although H3 is proficient in video editing, it falls short compared to Google's Gemini Omni Flash in text-to-video and image-to-video capabilities. Full model weights will be available shortly.
Key facts
- MiniMax released H3 video AI model
- H3 bests ByteDance's Seedance 2.0 in some tasks
- H3 is world's most powerful in video editing per Artificial Analysis
- H3 trails Google's Gemini Omni Flash in text-to-video
- H3 ranks behind Seedance 2.0 and Gemini Omni Flash in image-to-video
- H3 available via API and Hailuo AI platform
- MiniMax to open-source H3 weights under Community License
- Weights release slated for coming days
Entities
Institutions
- MiniMax
- ByteDance
- Artificial Analysis