Video AI: MiniMax challenges ByteDance with low price, open weights for new H3 model

MiniMax has introduced H3, its latest multimodal video generation model, in an attempt to challenge ByteDance's dominance in the field. The company claims H3 is currently the world's most powerful AI model for video editing, although it falls short of Google's Gemini Omni Flash in text-to-video tasks and is outperformed by both ByteDance's Seedance 2.0 and Gemini Omni Flash in image-to-video capabilities.

H3 is designed for commercial content creation in various sectors, including film, advertising, video games, and product design, and is currently accessible via an API and MiniMax's consumer-facing Hailuo AI platform. MiniMax plans to release the model's full weights soon, aiming to open up the ecosystem and allow users to customize their own versions using a wide range of AI hardware.

MiniMax's decision to open up H3's weights comes as closed-source models have long dominated the video generation space, with slower iterations and a less open ecosystem than large language models. The company will release the weights under the MiniMax Community License, which permits free non-commercial use and commercial use for organizations with under US$20 million in annual revenue, requiring attribution.

This licensing model is expected to become more prevalent among Chinese open-weight AI developers as they seek to commercialize their large models.

To stay competitive, MiniMax is offering H3 at a significantly lower price point compared to its peers. At 768p resolution, H3 costs US$7.80 per minute with audio, while ByteDance's Seedance 2.0 at 1080p costs US$22.45 per minute. While MiniMax's offering is pricier than Google's Gemini Omni Flash at US$6 per minute, the company is betting that its more accessible pricing will attract a wider range of users.