Back to models
MiniMax
Video modelNewHotMiniMax H3
MiniMax H3 is a general-purpose multimodal video generation model released by MiniMax. It interprets context from text, images, video, and audio together, then generates video with native stereo audio. For creators, the prompt does not have to carry every requirement: images can define subjects or composition, video can demonstrate movement and camera language, and audio can establish rhythm or sound direction. Start with text if you only have an idea. Image-to-video can use a starting image and an optional ending image, while reference generation can combine images, video, and audio. Official research materials highlight understanding complex instructions, rendering text and brand elements in the frame, and transferring motion from reference video. These capabilities suit product demonstrations, advertising drafts, interface concepts, and narrative clips with sound. Multimodal inputs offer more control but also require consistency between the materials. Define each reference's role first, then identify the elements that must be preserved in the prompt. Review generated text, logos, body movements, and audio-visual synchronization against the requirements of the work.
- Multimodal References
- Stereo Audio
Key features
Multimodal references
Combine text, images, video, and audio to specify subjects, movement, and sound direction.
Start and end frame guidance
Image-to-video uses a starting image and can include an ending image to plan the clip’s conclusion.
High resolution and native audio
Choose 768p or 2K output; the model supports generating audio together with the visuals.
