MiniMax H3 — the complete guide

MiniMax H3 is MiniMax's video generation model, available alongside Seedance 2.0 and 2.5 inside AI Cineworks — 2K resolution, reference-to-video generation across images, video, and audio, and per-second pricing. Here's what it can do and how to use it.

Core specs

What it's good for

Text-to-video

Generate a full clip from a text prompt alone — useful for concepting shots you don't yet have a source image or video for.

Image-to-video

Animate a still image into motion, with the output ratio automatically derived from the source image.

Reference-to-video

Attach up to 9 reference images, 3 reference videos, and 3 reference audio clips alongside a prompt to anchor the output to specific characters, locations, camera motion, or a soundtrack — the same OnSet workflow used with Seedance 2.0 and 2.5.

Pricing

MiniMax H3 is billed per second of output, based on resolution:

Cost scales linearly with clip duration — a 10-second 2K clip costs roughly 10× the per-second rate, with no separate resolution-based token formula to account for.

MiniMax H3 vs Seedance

MiniMax H3 sits between Seedance 2.0 and 2.5 on specs — its 2K resolution beats Seedance 2.5's 720p cap but falls short of Seedance 2.0's 4K, and its 15-second max clip length matches Seedance 2.0 exactly (Seedance 2.5 goes up to 30 seconds). Where MiniMax H3 differs most is pricing structure: simple per-second billing instead of a token formula tied to width × height × frame rate. See the Seedance 2.0 guide and Seedance 2.5 guide for the full comparison.

Frequently asked questions