Hailuo 03 video, in a workspace built for it

Hailuo3 AI runs MiniMax-H3 (Hailuo 03 / Hailuo 3.0): text-to-video, first- and last-frame control, and multimodal references. Output is native 2K with stereo audio, 4–15 seconds per clip — generate and iterate in the browser at hailuo3-ai.com.

Hailuo 03
Model
Hailuo 03 · MiniMax H3
Add multimodal references
Create now

What MiniMax-H3 does on Hailuo3 AI

H3 is MiniMax’s general-purpose multimodal video model. It reads text, images, video, and audio together, generates picture with native stereo sound, and supports instruction-based editing — so you can brief once and refine instead of restarting from zero.

00:15

Native 2K output

API resolution is 2K by default. H3 regenerates detail in-context rather than bolting on a separate upscaler, which helps small text and fine product detail survive the master.

Try it now
00:15

Native stereo audio on every clip

Official H3 generations include stereo audio with the picture — dialogue, effects, and atmosphere in one audiovisual pass, not a silent video you dub later.

Try it now
00:15

4–15 second clips (integer seconds)

Choose any whole-second length from 4 through 15. Long enough for a multi-beat hook; short enough to iterate without waiting on a full short film.

Try it now
00:11

Up to 12 multimodal references

In reference mode: up to 9 images, 3 videos (2–15s each, 15s total), and 3 audio clips (audio must pair with an image or video). Mix subject, motion, style, and voice in one brief.

Try it now
00:15

Instruction-based edits

H3 supports precise multimodal editing: change characters, objects, scenes, sound, or pacing with language instead of only regenerating from a blank prompt.

Try it now
00:15

UI, game, and stylized motion

Official demos cover product UI motion, game-style camera, and stylized animation — useful for previz, PV mood, and interface storytelling.

Try it now

How to generate Hailuo 03 on Hailuo3 AI

From account to export

Open Hailuo3 AI
STEP 01

Open Hailuo3 AI

Create an account at hailuo3-ai.com. The product is focused on MiniMax-H3 — you are not picking from a long multi-model menu.

Choose mode and inputs
STEP 02

Choose mode and inputs

Text-to-video needs a prompt and an explicit aspect ratio. Image-to-video uses first and/or last frames. Reference mode adds images, video, and optional audio under the official caps.

Set length, generate, export
STEP 03

Set length, generate, export

Pick 4–15 seconds at 2K, generate, preview with audio, then download the result or describe an edit instruction for another pass.

Built for Longer and Sharper Video

Ads and performance creatives

Ads and performance creatives

Generate sound-on variants for the same offer. Use reference images for product and talent so color and identity stay closer across takes.

Try it Now
Brand films and openers

Brand films and openers

Combine mood, character, product, and logo references in one multimodal brief — the pattern MiniMax highlights for premium brand films.

Try it Now
Short drama and character performance

Short drama and character performance

Vertical trailers and character-led beats benefit from face/voice references and multi-shot modeling inside a single 4–15s generation.

Try it Now
Product and e-commerce motion

Product and e-commerce motion

Native 2K masters for pack shots, feature demos, and listing media where still photos under-sell the product.

Try it Now

How to brief MiniMax-H3

Three things to set before you hit generate

01

Image 1

= face

Image 2

= product

Video 1

= camera

Audio 1

= voice

Label each reference

Tag every file with a role. H3 fuses image, video, and audio — vague labels waste generations.

02
Imagesmax 9 · 30 MB/file
Videosmax 3 · 50 MB/file
Audiomax 3 · 15 MB/file
Total files≤ 12

Audio needs an image or video pair — never alone.

Stay inside the caps

Stay under the official caps. Audio cannot go alone — always pair it with image or video.

03
Duration10s
4s15s

Aspect ratio

21:9
16:9
4:3
1:1
3:4
9:16

Lock length & frame

Pick 4–15 seconds. Text-to-video needs a ratio; image-to-video follows the frame; reference mode can stay adaptive.

Who Hailuo3 AI is for

If you already want H3, skip the multi-model maze

Need sound-on 9:16 and 16:9 variants of the same offer without configuring an API console.

Performance marketersPaid social

Want multimodal references (look, talent, product, logo) in one brief before a live shoot.

Brand & film teamsCampaign previz

Need 2K product motion that still holds on listing pages after export.

Commerce creativesPDP & marketplaces

Prefer instruction edits on a near-miss take over spending another full generation from scratch.

Short-form editorsChannels & reels

Use face and voice references so the same character can hold across short vertical beats.

Indie storytellersDrama & character

Prototype UI motion or game-style camera with stereo audio already in the draft.

Product & game designersUI / PV mood

Need sound-on 9:16 and 16:9 variants of the same offer without configuring an API console.

Performance marketersPaid social

Want multimodal references (look, talent, product, logo) in one brief before a live shoot.

Brand & film teamsCampaign previz

Need 2K product motion that still holds on listing pages after export.

Commerce creativesPDP & marketplaces

Prefer instruction edits on a near-miss take over spending another full generation from scratch.

Short-form editorsChannels & reels

Use face and voice references so the same character can hold across short vertical beats.

Indie storytellersDrama & character

Prototype UI motion or game-style camera with stereo audio already in the draft.

Product & game designersUI / PV mood

Need sound-on 9:16 and 16:9 variants of the same offer without configuring an API console.

Performance marketersPaid social

Want multimodal references (look, talent, product, logo) in one brief before a live shoot.

Brand & film teamsCampaign previz

Need 2K product motion that still holds on listing pages after export.

Commerce creativesPDP & marketplaces

Prefer instruction edits on a near-miss take over spending another full generation from scratch.

Short-form editorsChannels & reels

Use face and voice references so the same character can hold across short vertical beats.

Indie storytellersDrama & character

Prototype UI motion or game-style camera with stereo audio already in the draft.

Product & game designersUI / PV mood

FAQ

Model facts and the Hailuo3 AI product

Hailuo3 AI (hailuo3-ai.com) is a web product focused on MiniMax-H3 — the model also called Hailuo 03 or Hailuo 3.0. It is not the MiniMax official console; it packages H3 generation for creators who want a single-model workspace.

Every job needs a text prompt (up to 7,000 characters). Reference mode allows ≤9 images, ≤3 videos, ≤3 audio clips, ≤12 files total. Audio references must include an image or video. Per-file caps: image 30 MB; video 50 MB; audio 15 MB. Video reference clips are 2–15s each with total ≤15s.

Yes. MiniMax describes H3 as generating video with native stereo audio. Preview with sound on before you export from Hailuo3 AI.

MiniMax-H3 is MiniMax’s general-purpose multimodal video model (API model name MiniMax-H3). It accepts text, image, video, and audio context, outputs native 2K video with stereo audio, supports 4–15 second integer durations, and offers text-to-video, first/last-frame image-to-video, reference generation, and instruction-based editing.

Output resolution is 2K. Duration is any integer from 4 to 15 seconds. Aspect ratios: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16, or adaptive where the mode allows. Text-to-video requires a fixed ratio (not adaptive). First/last-frame image-to-video follows the input image framing.

Hailuo 2.3 (MiniMax-Hailuo-2.3) is a prior video line with shorter clips and up to 1080p on documented API settings (for example 6s/10s at 768P, 6s at 1080P). MiniMax-H3 targets native 2K, 4–15s, unified multimodal references (image/video/audio), native stereo audio, multi-shot modeling, and instruction-based editing in one general-purpose model.

Start on Hailuo3 AI

Create a free account, open the MiniMax-H3 workspace when generation is available on your plan, and upgrade if you need higher volume.

Create account

Pricing

Credit plans for MiniMax H3 (Hailuo 03). See how many 5s / 10s clips each tier covers.

Credits refill every billing cycle. Video estimates use MiniMax H3 at 5s / 10s · 2K.

Starter

Start experimenting with the magic of AI video.

$14/mo
Get Started
4,000 credits / month
  • Up to ~20 H3 videos (5s 2K)
  • Up to ~10 H3 videos (10s 2K)
  • MiniMax H3 · native 2K
  • Text / image / reference to video
  • 8 parallel generations
  • Watermark-free exports

Plus

Unlock deeper control and create at scale.

$34/mo
Get Started
12,000 credits / month
  • Up to ~61 H3 videos (5s 2K)
  • Up to ~30 H3 videos (10s 2K)
  • Add credits as needed
  • Commercial use rights
  • MiniMax H3 · native 2K
  • 16 parallel generations
  • Watermark-free exports
Most Popular

Pro

Create at full speed with uninterrupted flow.

$56/mo
Get Started
24,000 credits / month
  • Up to ~123 H3 videos (5s 2K)
  • Up to ~61 H3 videos (10s 2K)
  • Add credits as needed
  • Commercial use rights
  • Priority support
  • MiniMax H3 · native 2K
  • 32 parallel generations
Best Value

Wonder

Imagine without limits. Make your wonder.

$240/mo
Get Started
106,000 credits / month
  • Up to ~543 H3 videos (5s 2K)
  • Up to ~271 H3 videos (10s 2K)
  • Add credits as needed
  • Commercial use rights
  • Priority support
  • MiniMax H3 · native 2K
  • 32 parallel generations