MIMiniMax video model

Hailuo 3. The video model that takes direction.

MiniMax's open-weights model is built for control: instruction-guided edits, legible on-screen text and brand marks, and crisp 2K output with audio.

Try Hailuo 3->See all models

5 to 15 second clips
6 aspect ratios
Native audio
First and last frame control

The model

Most video models are great at generating and terrible at revising. Hailuo 3 (MiniMax H3) leans the other way: it is designed for precise, instruction-guided generation, where you say what should change and it changes that. It renders readable text and brand elements in the shot, holds a look between the first and last keyframe, and outputs at 2K with native audio, in clips from 5 to 15 seconds.

On Treza, Hailuo 3 sits in the same picker as Veo, Sora, and Kling. Use it in chat, or put it in a pipeline where a language model turns a brief into edit instructions and Hailuo renders them. Reference images bill at 4 cents each on top of the per-second rate, and like every model here it charges prepaid credits only on successful runs.

Instruction-guided generation

Built for controlled content generation and edits: describe the change you want and the model applies it rather than re-rolling the whole idea.

Text and brand rendering

Logos, signage, and on-screen words that stay legible. Rare among video models, and exactly what product and ad work needs.

2K output with audio

Sharp 2K frames with native audio generation, across six aspect ratios from 21:9 cinema to 9:16 vertical.

Keyframes at both ends

Pin a first and last frame and let the model travel between them, so a clip starts and lands exactly where the storyboard says.

Prompt to result

What to make with Hailuo 3

Real prompts, and what comes back. Every one runs in Treza chat or as a node on the canvas.

“A soda can rotates on a marble counter, brand name crisp on the label, morning light”

A product beat with the wordmark staying readable.

“Neon sign flickers on over a ramen shop, the shop name legible in the sign”

On-screen text rendered as part of the scene.

“A skateboarder carves an empty pool at golden hour, 21:9”

A cinematic wide in 2K.

“Same scene, but change the jacket to red and make it snow”

An instruction-guided revision instead of a fresh roll.

“A billboard timelapse over a city intersection, dusk to night”

Brand-space storytelling with legible signage.

“Macro shot of espresso pouring, steam curling, cafe logo on the cup”

Detail work with a brand element held sharp.

Beyond one prompt

Hailuo 3 is one node on a canvas

Chain it with language, image, and video models, then publish the whole pipeline as an API.

Step 01
Draft the brief
A language model node expands a one-line idea into a detailed, on-brand prompt before it ever reaches the model.
Step 02
Generate with Hailuo 3
Run variants in parallel, compare against other models on the same prompt, and keep every result in run history.
Step 03
Publish it as an API
The pipeline becomes a versioned endpoint. Your product calls it with the typed /invoke API or any OpenAI SDK.

FAQ

Hailuo 3, answered

What is Hailuo 3 best at?

Controlled work. When the brief has hard requirements, a logo that must read, on-screen text, a revision to an existing idea, or a clip that must start and end on exact frames, Hailuo 3 is the pick. For loose cinematic exploration, Veo 3.1 or Sora 2 Pro may wander more beautifully.

Is Hailuo 3 really open weights?

Yes, MiniMax released H3 as an open-weights model. On Treza you simply use it through the same picker and per-second pricing as every other model, with no hosting to manage.

How long and how sharp can clips be?

5 to 15 seconds per generation at 2K, in six aspect ratios with native audio. For longer pieces, stitch generations with the Sequence node or the timeline editor.

How much does Hailuo 3 cost on Treza?

Treza uses prepaid credits with no subscription. Each clip is charged at Hailuo 3's rate from your balance, and only successful generations are charged. Credit packs start at $5 and never expire.

Can I call Hailuo 3 through an API?

Yes. Build a pipeline that uses Hailuo 3 and publish it as a versioned HTTP endpoint. Call the typed /invoke API from your product, or point any OpenAI SDK at the OpenAI-compatible endpoint. If a better model ships later, swap it on the node without changing the API your product calls.