🐚MiniMax H3 Model - User Guide

This guide is continuously updated.

MiniMax H3: Open, General-Purpose Multimodal Video Intelligence

1. Model Overview

MiniMax H3 is a new generation of open, general-purpose multimodal video models. It marks a shift from specialized task models toward general multimodal intelligence. Instead of treating generation, editing, and reference-based creation across image, video, and audio as separate tasks, H3 understands text, images, video, and audio as one unified multimodal context—so it can interpret creative intent and produce more natural, coherent results. H3 is a step toward truly general-purpose multimodal intelligence.

2. Model Specifications

Core specification	MiniMax-H3
Model	MiniMax-H3
Output duration	5-15 seconds
Output aspect ratio	First/Last Frame: follows the aspect ratio of the uploaded image
Text-to-Video: supports user-selected 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16 aspect ratios
Omni Reference: supports user-selected 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16 aspect ratios, or Auto mode, which lets H3 determine the best ratio
Resolution	768p mode (coming soon)

For aspect ratios from 16:9 to 9:16, the short edge is 768 pixels. For wider formats, total resolution is approximately 1 megapixel—for example, 1536×672 at 21:9.
768p results can be upscaled to 1440p. 1440p mode
For aspect ratios from 16:9 to 9:16, the short edge is 1440 pixels. For wider formats, total resolution is approximately 3.7 megapixels—for example, 2976×1248 at 21:9. | | Frame rate | 24 FPS | | Audio output | Every generation includes native stereo audio. | | Input modes | | | First/Last Frame | - Images: 0, 1, or 2; dimensions: [256, 5760]; aspect ratio: 5:2 to 2:5
With no image input, H3 runs in Text-to-Video mode. | | Omni Reference | - Images: up to 9; dimensions: [256, 5760]
Videos: up to 3 clips; 2-15 seconds per clip; 15 seconds total; dimensions: [256, 5760]; aspect ratio: 5:2 to 2:5
Audio: up to 3 clips; 2-15 seconds per clip; 15 seconds total. Audio must be paired with at least one image or video and cannot be used on its own.
Mixed inputs: up to 12 files total
With no image, video, or audio input, H3 runs in Text-to-Video mode. | | Supported input formats | Video: H.264/AVC, H.265/HEVC; embedded audio: AAC, MP3 Image: JPG, JPEG, PNG, WEBP, HEIC, HEIF Audio: WAV, MP3 | | File size limits | Video: 50 MB per file; image: 30 MB per file; audio: 15 MB per file. Limits apply per file, not to the combined upload size. API request body: 64 MB. For API workflows, URL-based media input is recommended. | | Prompt length | Up to 7,000 characters |

3. Key Capabilities & Examples

Capability	What it does
Native multimodal understanding and generation	Accepts text, image, audio, and video inputs; understands subjects, motion, sound, emotion, camera language, visual style, and creative intent; and blends multiple references into a complete audiovisual result.
Precise multimodal editing and control	Edits people, objects, scenes, sound, and pacing while following detailed instructions. Creators can keep iterating on existing content and make targeted changes to visuals, audio, and story details.
Production-ready content across commercial use cases	Built for film, advertising, brand, e-commerce, and gaming workflows. H3 handles on-screen text, subtitles, brand assets, creative effects, product showcases, UI/UX motion demos, game visuals, and stylized content—helping teams move faster from concept and visual pitch to production.

Production-Ready Content Across Commercial Use Cases

As video models become more capable, they can support a much broader range of real-world production workflows.

H3 can render text, subtitles, brand assets, UI/UX, game content, and e-commerce visuals across film, advertising, gaming, brand, and retail use cases.

Use it to explore short-form concepts, test visual directions, create more expressive product showcases, bring IP characters into new actions and settings, and support concept validation, storyboard previews, and visual pitches.

Brand Films & Cinematic Content

Trailers, TV commercials, and premium brand films

BOUND

Visual Concepts & Motion Design

Creative shorts, VFX packages, music visuals, visual experiments, trend-led content, and social assets