Models · 31 July 2026 · 1 min read
MiniMax H3: video with sound, from many kinds of input
Launched on 31 July 2026: one model that reads text, pictures, video and sound together, and returns video with stereo audio.
By The Crewnema team
MiniMax launched H3 (also called Hailuo 3.0) on 31 July 2026. MiniMax describes it as a single model that understands text, images, video and audio together, rather than a video model with other tools attached. This is what the company said at launch.
What MiniMax announced
- Video of up to 15 seconds at 2K resolution.
- Native stereo sound, with voice, sound effects and music generated together with the picture and not added afterwards.
- Reference and editing across kinds of input: image to image, image to video, audio to audio, and audio and video together.
- A plan to release the model’s weights openly “in the coming days”, subject to applicable laws, which had not yet happened at the moment of the announcement.
What it means for an advert
Sound is half of an advert, and a model that makes voice, effects and music together with the picture is easier to keep in sync than one where sound is added afterwards. Editing by reference is useful for the same reason as in other recent models: you can show the model what you want kept and change the rest.
In Crewnema
MiniMax H3 Max is the video model on the free plan. We picked it for the free level because it is among the lowest-cost options on the roster, not because it is the best looking: in our own early tests its clips were noticeably grainy, which is also why it is not the first rung of the paid levels. It is good for checking an idea, a script and the framing end to end. If you want the finish, move up a level. See the four quality levels.
Sources
Make your first video
Crewnema is in private beta and lets businesses in a few at a time. Request access and we will write when yours is open.