MiniMax H3

Name: Cosmic crosswalk transformation
Uploaded: 2026-09-15T00:00:00+08:00
Duration: 15 s
Description: A MiniMax H3 reference-to-video demo that keeps one character consistent from a night skyline through a statue tunnel and a cosmic passage to a daylight Manhattan crosswalk.

API Docs Start creating

Multimodal VideoNative Stereo Audio2K Generation

MiniMax H3 is a general-purpose multimodal video model that combines text, image, video, and audio context in one generation workflow. It creates videos up to 2K and 15 seconds with native stereo sound, while supporting first-and-last-frame control, mixed-media references, motion transfer, multi-shot storytelling, and instruction-guided video editing.

Inputs

Text / Image / Video / Audio

APIXO Price

$0.09–$0.13 / second

Resolution

768P / 2K

Duration

5-15 Seconds

Audio

Native Stereo Sound

Capabilities

MiniMax H3 Capabilities

Unified Context

Combine written direction with image, video, and audio references in one request. Assign each source a role such as appearance, motion, camera, voice, rhythm, environment, or style.

First & Last Frames

Define an opening frame, an ending frame, or both to guide transformations, transitions, product reveals, and storyboard beats while preserving a clearer visual trajectory.

Mixed-Media References

Use reference images, videos, and audio together to communicate character identity, product cues, movement, camera language, vocal performance, music, or ambience.

Native Stereo Sound

Generate voice, sound effects, ambience, and music together with the video as native stereo audio rather than assembling isolated audio tracks after generation.

Native Multi-Shot

Build short sequences with coordinated shot changes, scene progression, action continuity, and audiovisual pacing from one prompt-led or reference-guided workflow.

Editing & Motion Transfer

Transform existing visual material through natural-language direction or transfer motion from video references while reviewing identity, composition, occlusion, and source-detail fidelity.

Use Cases

Production concepts suited to MiniMax H3

Campaign Production

Audiovisual Product Stories

Combine product imagery, written direction, motion references, and sound intent to develop campaign concepts. Review labels, logos, materials, dimensions, and product geometry before approving generated commercial assets.

Character Direction

Reference-Led Performances

Establish a recurring subject with visual references, then guide movement, camera behavior, voice, or scene context through additional media. Evaluate identity continuity, anatomy, expression, and performance timing across the complete result.

Creative Development

Narrative Previsualization

Translate a treatment into short scenes that combine action, camera language, ambience, and audiovisual pacing. Use generated footage to evaluate ideas before committing to physical production or detailed post-production.

Guided Revision

Footage Transformation

Explore style, environment, subject, or scene changes through natural-language instructions and optional references. Inspect temporal boundaries, occlusion, unintended modifications, and source-detail preservation frame by frame.

Specifications

MiniMax H3 Model Specifications

Confirmed against the live APIXO deployment.

Text / Image / Video / Audio

Input

T2V / I2V / Reference

Workflows

768P / 2K

Resolution

5–15 Seconds

Duration

Native Stereo

Audio

6 Ratios + Adaptive

Framing

Notes & FAQ

MiniMax H3 availability and integration guidance

Notes

MiniMax H3 runs on the APIXO asynchronous task endpoint. Submit to generateTask and poll statusTask, or supply a callback URL for webhook delivery.

Three modes are exposed: text-to-video, image-to-video, and reference-to-video. Image-to-video takes 1-2 images as the first and optional last frame; reference-to-video needs at least one image or video reference.

Output is 768p or 2K at 5-15 seconds, whole seconds only. Resolution sets the per-second rate. Prompts are limited to 4,000 characters. Reference video and audio are each capped at 3 files, 2-15 seconds per file and 15 seconds in total.

“MiniMax H3” is the APIXO model ID. Do not carry over parameter values, quality tiers, or billing behavior from Hailuo 2.3 or Hailuo 2.3 Fast — they are separate routes with different contracts.

References guide generation but do not guarantee exact identity, typography, logos, materials, or product geometry.

Review anatomy, hands, object contact, rapid motion, occlusion, camera transitions, speech, and audiovisual synchronization before release.

Frequently Asked Questions

Yes. The model is live on the APIXO asynchronous task endpoint and can be called with the same API key as every other model in the catalog. Generation is also available in the Playground on this page.

Text-to-video, image-to-video, and reference-to-video. Reference-to-video accepts up to 9 images, 3 videos, and 3 audio files as combined creative context; audio cannot be supplied on its own.

768p or 2K output at 5 to 15 seconds, in whole seconds. 2K is the default and the highest tier the route exposes; 768p bills at a lower per-second rate. Aspect ratio can be set to 21:9, 16:9, 4:3, 1:1, 3:4, or 9:16 for text-to-video and reference-to-video; image-to-video takes its framing from the supplied first frame.

Output video is billed per second in every mode, at the rate that matches your chosen resolution — 768p is cheaper per second than 2K. In reference-to-video, the actual seconds of your reference videos are billed at the same per-second rate, and the first 5 reference images are free with each additional image charged separately. Reference audio is not billed. See the pricing tab for the current rates.

Yes. Audio is produced in the same pass as the picture as native stereo, with speech, sound effects, ambience, and music modelled jointly rather than as separate tracks.

No. Mixed-media references communicate creative intent but do not guarantee exact identity, motion, voice, product structure, or frame-level reproduction. Production use requires visual and audio review.