Video Gen Editor

MiniMax H3: Cinematic AI Video Up to 1440p

MiniMax H3 is the full version of MiniMax's H3 video model, aimed at cinematic, multi-shot work. It renders at up to 1440p, generates sound natively, and takes images, videos and audio as references, so a character's face, voice and movement can stay consistent from one shot to the next.

Free sign-up · see the cost before you generate

MiniMax H3 specs

Made by
MiniMax
Clip length
5 to 15 seconds
Resolutions
768p, 1440p
Shapes
16:9, 4:3, 1:1, 3:4, 9:16, 21:9
Starts from
Text, Image, Video, Audio
Extra inputs
First and last frames, up to 9 reference images, up to 3 reference videos, up to 3 reference audio clips
Sound
Generated with the video

MiniMax H3 pricing

MiniMax H3 price per resolution
ResolutionPer second5-second clip10-second clip
768p$0.08$0.40$0.80
1440p$0.13$0.65$1.30

These are the published per-second rates. The composer shows the exact estimate for your settings before you generate, and failed generations aren't charged. Prices last checked September 24, 2026. Credits come with our plans.

What it's good at

  • 768p or 1440p output, the sharpest MiniMax tier.
  • Up to 9 reference images, 3 reference videos and 3 audio references.
  • Native, synchronized sound rather than a separate dubbing pass.
  • First-frame and keyframe-guided generation.
  • Can continue an existing clip or audio segment, and edit from instructions.

Best for

  • Character-driven scenes

    Keep a face and a voice consistent across shots.

  • Cinematic shots

    Film-like motion and framing at 1440p.

  • Music videos

    Audio references alongside an artist photo, for the AI music video generator.

  • Continuations

    Pick up a clip where it ended, with the same voice.

Things to know

  • Clips are 5 to 15 seconds.
  • The first 5 reference images are included; each one after that adds $0.04. Reference video adds $0.13 per second.
  • Frame images can't be mixed with reference images, videos or audio.
  • With frame images, use a resolution preset rather than an exact size.

A MiniMax H3 prompt to try

Close-up of an old jazz trumpeter on a smoky club stage, a single spotlight, he lifts the trumpet and plays a slow phrase, the crowd murmuring in the dark.

Add a reference image of the musician, and a short voice or audio reference, to keep him consistent in later shots.

Alternatives to MiniMax H3

Compare every AI video model

Use MiniMax H3 for

Frequently asked questions

What's the difference between MiniMax H3, H3 Max, H3 Max Turbo and H3 Fast?

H3 is the full model, with references and output up to 1440p. H3 Max and H3 Max Turbo are faster, cheaper versions that work from text or frames only. H3 Fast is a quick 480p model that keeps reference support.

Does MiniMax H3 generate sound?

Yes, natively and in sync with the picture. It can also take audio references to keep a voice consistent.

How much does MiniMax H3 cost?

From $0.08 a second at 768p to $0.13 at 1440p. A 5-second 768p clip costs about $0.40.