Founded in early 2022, MiniMax is driven by the mission to 'co-create intelligence with everyone,' dedicated to advancing the frontiers of AI and achieving Artificial General Intelligence (AGI). MiniMax has independently developed a series of multimodal foundation models with powerful code and Agent capabilities, as well as ultra-long context processing, capable of understanding, generating, and integrating multiple modalities including text, audio, image, video, and music.
This is the 10th launch from MiniMax. View more
MiniMax H3
Launching today
MiniMax H3 is an open multimodal model that generates 2K video with native stereo sound. It unifies text, image, and audio inputs, excelling at accurate text rendering, visual packaging, and complex instruction following for commercial content creation.
Free Options
Launch Team
Flowtica Scribe
Hi everyone!
@MiniMax H3 is especially good at turning a mixed set of references into finished-looking motion work.
You can mix text, images, video, and audio in one request, then simply tell H3 what you want to borrow from each reference. It can follow the same character, camera movement, voice, or overall visual style and turn everything into a 2K video with native stereo sound.
This makes H3 especially useful for commercial creative work. The output can feel much closer to a finished piece, with the typography, motion, pacing, and sound working together across product videos, motion posters, music visuals, and ecommerce campaigns.
The API is live now, and the weights are coming!
Text rendering is the claim I'd want tested hardest here, because a motion poster lives or dies on one word being right and video models have historically turned typography into soup. The useful test isn't whether it renders clean once, it's whether you can swap that word for a longer one and get the same layout back. Everything in a branding workflow is a re-render, so consistency across takes matters more than any single take. Native stereo in the same pass is the part that actually removes a handoff.