MiniMax H3: Overview

What Is MiniMax H3?

MiniMax H3 is a multimodal video model available in Luma Agents. It's built to work with more than a visual prompt: it can take text, images, video, and audio together and generate a unified audiovisual result in a single pass, matched video and sound included. That means one generation can combine a character's look, motion from a reference video, a voice, camera direction, dialogue, ambience, and music all at once.

Think of H3 less as a video generator and more as an audiovisual scene generator. Instead of only describing what a video should look like, it's built to take direction on what the finished scene should be like to watch and hear.

What Makes It Different

Most video workflows split the work into separate steps: generate the picture, find a voice, then mix them together. Every handoff between those steps is a place where things like lip sync can drift, and fixing it usually means starting over.

H3 closes that gap by bringing all the signals into one shared generation. Text, image, video, and audio references go in together, and dialogue, sound effects, and ambience come out already timed to the action on screen.

What You Get

• Native 2K resolution at 24fps, in clips of 5 to 15 seconds

• Dialogue and sound generated in the same pass, timed to the action

• Image, video, and audio references combined for character, style, motion, and voice consistency

• Direction over subject, on-screen text, camera behavior, and motion, by instruction

• Video-to-video motion transfer

For guidance on how to prompt H3 and choose the right inputs, see MiniMax H3: Best Practices.

Where It's Useful

H3 is a strong fit whenever a video depends on several kinds of creative direction working together, such as character-driven scenes, commercials, product films, stylized content, music-driven work, narrative clips, motion references, dialogue-heavy scenes, and any result where sound matters as much as picture. Luma's and MiniMax's own examples span brand films, social creative, short-form narrative, product advertising, UI motion, games, and stylized animation.