MiniMax H3: 2K AI Video Generator with Native Audio

Generate 5 to 15 seconds at 2K and 24fps from a text prompt, an image, keyframes, or multimodal references. The current browser workspace supports up to 9 image references plus one video and one audio reference.

No card required. First generation is free.

15smax lengthOne continuous generation
2K · 24fps1440px short edgeDetail ready for the edit
Native stereoscore, dialogue, foleyRoom tone timed to picture
9 + 1 + 1workspace referencesImages, video, and audio
MODEL OVERVIEW

What Is MiniMax H3?

MiniMax H3 is a next-generation AI video Generator that understands text, images, video, and audio in one connected workflow.

Omni-Modal Video Generation

MiniMax H3 is MiniMax's general-purpose omni-modal video generation model. It reads text, images, video, and audio in one shared context rather than treating each input as a separate task. That keeps the MiniMax H3 workflow direct: write the intention, attach the references that matter, and let the model connect them in a single video.

Production-Ready Video Creation

For production work, MiniMax H3 generates 5 to 15 seconds at up to 2K with native stereo audio. A single prompt can carry camera direction, dialogue, sound cues, and visual constraints, so a brief describes not just what appears on screen but how the shot moves and resolves. That makes MiniMax H3 practical for advertising concepts, product films, social creative, game cinematics, and previsualization.

Reference-Led Creation & Keyframes

MiniMax H3 is built for reference-led work as much as prompt-led generation. A character image establishes identity, a clip communicates motion or camera language, and a recording guides voice or rhythm. First-and-last-frame generation adds a clear visual destination for product reveals, transformations, storyboard transitions, and controlled camera moves. MiniMax released H3-Base open weights on August 3, 2026 for eligible open-source and customized workflows.

T

Text

Describe the scene, movement, camera, lighting, dialogue, and sound.

Images

Guide character identity, product appearance, style, or composition.

Video

Reference motion, performance, camera behavior, and scene timing.

Audio

Guide voice, rhythm, ambience, music, or sound direction.

Four Ways to Generate Videos with MiniMax H3

The current MiniMax H3 workspace offers four clear starting points: create from text, animate one image, control a transition with first and last frames, or guide the result with multimodal references.

Text to Video

Describe the complete shot, including the subject, action, environment, camera movement, lighting, visual style, and sound. MiniMax H3 turns your creative brief into a detailed video at up to 2K resolution.

Choose from six MiniMax H3 aspect ratios, ranging from cinematic 21:9 to vertical 9:16, for advertising, social media, film concepts, and product content.

Generate Text to Video

Image to Video

Upload a product photo, portrait, illustration, storyboard frame, or campaign visual and describe how the subject, camera, environment, and sound should develop.

The source image establishes the opening composition while the prompt directs motion, timing, atmosphere, and the final moment.

Animate an Image

First and Last Frame

Provide both an opening and final frame to define how the scene begins and ends. MiniMax H3 generates the motion and transition between the supplied images.

Use this workflow for product transformations, visual reveals, storyboard animation, scene transitions, and videos that need a controlled final frame.

Generate from First and Last Frames
Reference image for the generated video
Audio reference

Reference to Video

Combine images, a video clip, and supported audio to guide character identity, product appearance, movement, camera behavior, voice, sound, and timing.

The current browser workspace accepts up to nine image references plus one video and one audio reference. Give every uploaded asset a clear role in the prompt.

Use this workflow when text alone cannot communicate the identity, product design, motion, or audiovisual direction you need.

Generate from References

MiniMax H3 Video Examples

Explore MiniMax H3 prompts, motion direction, and visual approaches created with MiniMax H3.

01

Studio product

A brushed metal watch turns slowly under controlled strip lighting. Macro lens, black cyclorama, quiet mechanical foley.

Use this direction
02

Night drive

Low tracking shot beside a car moving through wet neon streets. Reflections remain stable while the camera accelerates.

Use this direction
03

Editorial portrait

A portrait subject turns toward window light as the camera makes a restrained handheld push-in. Preserve identity and wardrobe.

Use this direction
04

Architectural light

Sunlight travels across a concrete interior from morning to dusk in one composed shot, with subtle room tone.

Use this direction
05

Fashion motion

Full-body runway movement with fabric physics, a lateral tracking camera, and a clean stereo score timed to each turn.

Use this direction
06

Coastal aerial

Aerial camera descends through morning mist toward a dark coastline. Natural wind, distant surf, steady horizon.

Use this direction
07

Glass and water

High-speed product shot of water wrapping around clear glass, then settling into a sharp final pack frame.

Use this direction
08

Game cinematic

A lone explorer crosses an illuminated industrial chamber. Wide lens, controlled orbit, layered machine ambience.

Use this direction
09

Performance cut

A singer performs in a compact rehearsal room. Match lip movement, voice character, camera rhythm, and practical lighting.

Use this direction

How to Create a Video with MiniMax H3

  1. 01

    Write the shot

    Tell MiniMax H3 who or what is in frame, what changes during the clip, how the camera moves, and how the scene should sound. Use clear constraints for details that must stay stable.

  2. 02

    Attach references

    Add only the images, clips, or audio that sharpen the direction. MiniMax H3 can use references for character identity, product detail, motion, camera language, voice, and pacing.

  3. 03

    Generate and download

    Review the complete MiniMax H3 clip for subject consistency, motion, camera behaviour, audio, and the final frame. Refine one instruction at a time, then download the version that works.

MiniMax H3 Capabilities

CapabilityHailuo 2.3MiniMax H3
OutputEarlier Hailuo workflowMiniMax H3: up to 2K at 24fps
LengthShort-form generationMiniMax H3: 5–15 seconds
AudioSeparate production stepMiniMax H3 native stereo generation
ContextLimited creative inputsMiniMax H3 reads text, images, clips, and audio together
KeyframesSingle opening imageMiniMax H3 first-and-last-frame control
WeightsHosted workflowMiniMax H3-Base open weights released on August 3, 2026

Choose the MiniMax H3 workflow that gives your shot the clearest creative direction.

MiniMax H3 Credit Pricing

Choose a one-time credit pack for text-to-video, image-to-video, first-and-last-frame, and reference-based creation. Start small to test your workflow or save more with higher-volume credits.

One-time payment · No subscription · No auto-renewal · Credits never expire

Credit Pack

Starter

$9.90 one-time

$0.050 per credit

200 credits

8 standard 5s 2K videos

$1.24 per standard video

A low-cost way to test MiniMax H3 prompts, images, references, and creative settings before purchasing a larger pack.

  • A standard 5-second 2K generation uses 25 credits
  • Access text-to-video, image-to-video, and reference workflows
  • Create up to 2K videos with native audio in supported modes
  • Credit use varies with duration, resolution, and reference inputs
  • Failed or rejected generations are automatically refunded
  • Credits never expire

Best Value

Studio

$99.90 one-time

$0.040 per credit

2,500 credits

100 standard 5s 2K videos

$1.00 per standard video

Built for production sprints, advertising campaigns, client projects, and teams generating videos at a higher volume.

  • A standard 5-second 2K generation uses 25 credits
  • Get the lowest cost per credit
  • Credit use varies with duration, resolution, and reference inputs
  • Use all available MiniMax H3 generation workflows
  • Ideal for campaigns, product content, and production testing
  • Failed or rejected generations are automatically refunded
  • Credits never expire

MiniMax H3 credits are only spent on a successful generation. Failed or rejected jobs are refunded automatically.

MiniMax H3 FAQ

What resolution and video length does MiniMax H3 support?

MiniMax H3 can generate videos from 5 to 15 seconds long at up to 2K resolution and 24 fps. Available MiniMax H3 settings can vary by generation mode and the current service interface.

What inputs can I use with MiniMax H3?

MiniMax H3 accepts written prompts, images, video clips, and audio references. In MiniMax H3, use a prompt for the creative direction and add references when you need greater control over identity, appearance, motion, camera language, voice, or timing.

Does MiniMax H3 generate native audio?

Yes. MiniMax H3 generates native stereo audio with the video. Describe the intended dialogue, ambience, music, or sound effects in the MiniMax H3 prompt, then review the finished clip to make sure the sound supports the picture.

Can MiniMax H3 create a video from text?

Yes. Describe the subject, setting, action, camera movement, lighting, style, sound, and final moment. A focused MiniMax H3 prompt that explains what changes during the MiniMax H3 shot is usually more controllable than a list of unrelated keywords.

Can I use first and last frames with MiniMax H3?

Yes. Upload a starting image, an ending image, or both to guide how a MiniMax H3 video opens and concludes. This MiniMax H3 workflow is useful for product reveals, storyboard animation, controlled scene transitions, and shots with a precise final composition.

Can MiniMax H3 use image, video, and audio references?

Yes. MiniMax H3 can combine reference images, clips, and audio with a prompt. Use only the assets that clarify the MiniMax H3 result: an image can lock a character or product, a clip can show motion, and audio can suggest voice or pacing.

How many references can I upload in this workspace?

The current browser workspace accepts up to nine image references plus one video and one supported audio reference. Give each asset a clear role in the prompt and use only the references that directly support the intended shot.

Which aspect ratios are available in MiniMax H3?

Supported MiniMax H3 workflows offer 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16. Choose a MiniMax H3 widescreen format for websites and cinematic concepts, square formats for feed posts, and vertical formats for short-form mobile video.

Can MiniMax H3 create readable text and brand elements?

MiniMax identifies accurate text and brand presentation as a MiniMax H3 capability. For subtitles, signage, credits, labels, or interface elements, keep the wording concise and inspect the generated MiniMax H3 frame before publishing.

How do I write a better MiniMax H3 prompt?

Start with the subject and environment, then explain the visible action over time. Add one clear camera instruction, the lighting and visual style, the intended sound, and any details MiniMax H3 should keep unchanged. Revise one MiniMax H3 instruction at a time when refining a result.

Create Your First MiniMax H3 AI Video

Turn a prompt or approved keyframe into a 2K · 15 second MiniMax H3 video with native stereo audio.

Generate a videoNo card required · Start with a prompt