MiniMaxAI released MiniMax-H3 on Hugging Face
MiniMaxAI released MiniMax-H3 on Hugging Face. It is an image-text-to-video model with license 'other' and library 'minimax-h3'. It supports output durations of 4–15 seconds, resolutions up to 2K (via H3-Regenerate-2K), 24 FPS, 32 kHz stereo audio, and 11 languages. Input modes include H3-Base-FL2VA (first-and-last-frame) and H3-Base-Ref2VA (omni-reference with up to 9 images, 3 video clips, 3 audio clips, or 12 mixed files).
China context
- Original name
- 稀宇科技
- Outside China
- Open weights · huggingface.co
- Claims
- Company-reported; not yet independently evaluated
- For builders
- Builders outside China can access the model weights on Hugging Face and integrate video generation into their applications, but must comply with the 'other' license terms.
- For investors
- MiniMax's open-weights release signals a strategy to compete globally by fostering a developer ecosystem, which may increase demand for its API and platform services.
MiniMaxAI released MiniMax-H3 on Hugging Face, an omni-modal generative system that supports unified understanding of text, images, video, and audio, and can generate video with native stereo audio at up to 2K resolution and 15 seconds duration.
The system consists of three modules: H3-Context-IR (hosted preprocessing and orchestration, not open-sourced), H3-Base (generates 768p audio/video), and H3-Regenerate-2K (upscales to 2K). H3-Context-IR converts multimodal inputs into a Context Intermediate Representation for generation.
MiniMax is releasing open-weights models on Hugging Face, making advanced video generation capabilities accessible to developers outside China. The model supports a wide range of input modes and languages, positioning it for global adoption.
Developers can use MiniMax-H3 via API (global: platform.minimax.io, CN: platform.minimaxi.com) or app (global: hailuoai.video, CN: hailuoai.com). The open-weights release may drive adoption of MiniMax's platform services.
Observable next signals include whether MiniMax releases H3-Context-IR or provides API access to it, and whether developers adopt the model for production video generation tasks.