MiniMax H3 resourcesBuild better video promptsOpen generator

Model introductionMiniMax H3 · Independent creative experience

A deliberate workspace for directing moving images.

MiniMax H3 brings several established MiniMax video-generation patterns into one focused interface. The product is independent; its factual capability map is grounded in MiniMax’s officially released Hailuo 2.3, Hailuo 02, and S2V-01 documentation.

Try the browser demo Read the API map

Official capability signal

Hailuo 2.3 advances motion, physical behavior, and expressive detail.

MiniMax describes improvements in complex body movement, object motion, command following, stylization, and live-action micro-expressions.

Motion: Complex body movement

Expression: Subtle facial performance

Style: Live action to illustration

Direction: Stronger command response

Creative inputs: Text · Image · Frames · Subject

Public flagship: MiniMax-Hailuo-2.3

Output window: 6s or 10s

Maximum listed output: 1080P · 6s

Capability map

What the public Hailuo 2.3 release emphasizes.

The official release focuses on controllable movement and expressive video performance rather than one headline resolution number.

Complex human movement

Hailuo 2.3 is presented as more fluid, natural, precise, and controllable when rendering demanding body motion.

Physical and object behavior

The release highlights stronger understanding of physics and improved response to motion commands for both people and objects.

Facial micro-expression

Subtle live-action expression changes are described as more natural, supporting character-led close shots and performance.

Wider stylization

The general model extends support across anime, illustration, ink-wash aesthetics, game CG, and live-action looks.

Camera-aware direction

Prompts can combine natural language with documented bracketed commands for explicit pans, pushes, tilts, tracking, and more.

Fast image workflow

MiniMax-Hailuo-2.3-Fast is a separate image-to-video model intended to reduce generation time and cost for repeated production.

Workflow ownership

Four starting points, mapped to their real model families.

The MiniMax H3 interface groups these modes for creative continuity. The official API does not assign every mode to the same model.

Starting point	What you provide	Documented model family	Best for
Text to video	Prompt	MiniMax-Hailuo-2.3	Inventing a scene from written direction
Image to video	First frame + optional prompt	Hailuo 2.3 or 2.3 Fast	Animating a controlled visual opening
First and last frame	Two images + optional prompt	MiniMax-Hailuo-02	Directing a transition toward a chosen ending
Subject reference	Character image + prompt	S2V-01	Prioritizing facial identity consistency

Creative system

A clear source, a directed prompt, and an asynchronous result.

The product experience is organized around decisions a creator can control before generation begins.

Source

Choose how the shot begins

Start from pure text, a first frame, a first-and-last-frame pair, or a subject reference depending on what must remain controlled.

Text · Image · Frames · Subject

Direction

Write the movement, not just the look

Describe the subject’s action, scene evolution, camera path, lighting, atmosphere, and intended final beat.

Subject + scene + motion + camera + mood

Settings

Match duration to shot complexity

Use a supported resolution and duration. Keep complex camera choreography within a short, coherent action window.

6s / 10s · 768P / 1080P

Result

Design for a queued workflow

The official API returns a task ID before a file is ready. Product interfaces should make waiting, recovery, and final retrieval explicit.

task_id → file_id → download_url

Output constraints

Duration and resolution are a model-specific pair.

For the public Hailuo 2.3 text and image workflows, a longer clip trades away the 1080P option.

Model	Mode	6 seconds	10 seconds
MiniMax-Hailuo-2.3	Text · Image	768P or 1080P	768P
MiniMax-Hailuo-2.3-Fast	Image	768P or 1080P	768P
MiniMax-Hailuo-02	Text · Image · First/last	768P or 1080P	768P

These combinations reflect the current public API reference. Always re-check official documentation before making a permanent product promise.

Where it fits

Short shots with a clear visual objective.

The strongest workflows give one clip a focused job, then assemble multiple results through a larger editing or campaign process.

Previsualization

Film and story concepts

Explore camera paths, blocking, light, and atmosphere before committing to a physical shoot or full edit.

Campaign production

Product and social motion

Turn a controlled product frame into short movement studies, ad variations, or visual hooks.

Character performance

Expressive close shots

Direct one readable action and a subtle emotional change rather than several competing story beats.

Stylized exploration

Animation and art direction

Test illustration, anime, ink-wash, game-CG, and cinematic live-action directions with concise motion language.

Current site boundaries

What the MiniMax H3 demo does—and does not do.

The current build is intentionally transparent about its simulated generation experience.

Local preview

Reference files stay in the browser

Selected files are used only for temporary previews. This implementation does not upload them to a generation provider.

No API request

Generation is simulated

The queue and result flow demonstrate product behavior without creating an official MiniMax task.

No payment

Credits and pricing are presentation data

No checkout, subscription, wallet debit, or real account balance is connected to the current interface.

No endorsement

Independent product experience

The site is not affiliated with or endorsed by MiniMax. Official names are used only to explain public capabilities accurately.

Model questions

A precise way to describe the experience.

Is MiniMax H3 the same model as MiniMax-Hailuo-2.3?

No. MiniMax H3 is the product name used by this independent site. MiniMax-Hailuo-2.3 is a real public model ID documented and released by MiniMax.

Why does the interface show four input modes?

The four modes represent workflows documented across the MiniMax video API family. Text and image generation are available on Hailuo 2.3, first-and-last-frame generation is documented for Hailuo 02, and subject reference uses S2V-01.

Does Hailuo 2.3 support 10-second 1080P output?

The current public reference lists 1080P for 6-second Hailuo 2.3 output. The 10-second option is listed at 768P.

Are the videos shown here official Hailuo 2.3 outputs?

No such claim is made. The current site uses presentation media and a client-side simulation to demonstrate a creative workflow.

Start with a focused shot

Choose the source. Direct the movement.

Use the browser demo to frame an idea, then carry the prompt language into a supported production workflow.

Open the generator Learn prompt structure Read the Hailuo 2.3 release