Hailuo 3 Max API and Pricing, Explained Through a Real Pipeline

Hailuo 3 Max is MiniMax's further-trained version of Hailuo 3, usually written as H3 Max. It renders clips of 5 to 15 seconds from a text prompt, from an exact first frame, or between a first and a last frame you supply, at 480p or 768p in six aspect ratios. It reached OpenRouter at the start of September 2026, and it joined Treza's video catalog on October 4.

This post covers what it costs at each resolution, how it compares with Hailuo 3 on price and inputs, what the endpoint accepts and what it leaves out, and how to call Hailuo 3 Max as a published API endpoint. Every parameter and price below is read off Treza's live video model catalog as of October 2026, and every behavior described was checked with real renders.

What you get before you talk about price

The row that matters is keyframes. Hailuo 3, the model this one was trained from, takes neither a first nor a last frame on the endpoint we run. Hailuo 3 Max takes both, so a shot can start on a still you already have and land on another.

The other property is speed. In our launch-week test renders, five second clips came back in 12 to 14 seconds, at both resolutions and with or without a first frame. Capability detail lives on the Hailuo 3 Max model page.

How Hailuo 3 Max bills

Per second of output, at a rate set by the resolution:

A first frame or a last frame does not change the rate. Our 5 second test clip at 480p cost 25 cents of provider cost from a prompt, and 25 cents again with a first frame attached. The same 5 seconds at 768p, the default, cost 40 cents. The longest clip the model renders, 15 seconds at 768p, is $1.20.

For scale, HeyGen Video starts at 2 cents a second at 480p, Wan 3.0 and Grok Imagine at 5 cents, and Veo 3.1 Lite at 3 cents a second for a silent clip at 720p.

Hailuo 3 Max vs Hailuo 3: two cents a second for keyframes

Same vendor, same price at 480p, and the two models split on three things.

At 768p, the default on both nodes, Hailuo 3 Max costs two cents a second more than Hailuo 3. The two cents buy control over where a shot starts and ends, and give up the 2K and 4K ceiling and the generated soundtrack.

That makes the choice simple. A shot that has to open on your product photo, or move from a sketch to a finished render, goes to Hailuo 3 Max. A shot that will be shown large, or that should arrive with its own ambience, goes to Hailuo 3, which the Hailuo 3 pricing post covers in the same detail. When in doubt, run one prompt through both and compare.

First and last frames: what they do to a shot

A first frame pins the opening image exactly, and the clip animates out of it. That is the technique the first frame post walks through, and the image to video generator covers it from the still's side. Add a last frame and the model fills in the motion between the two stills, so the shot lands where you decided.

Two observations from our renders. With a first frame attached and no aspect ratio set, the clip took the shape of the still: our square test image came back as a 480 by 480 clip. Set the aspect ratio explicitly whenever the still and the placement disagree. And a first frame bills at the plain rate, so there is no pricing reason to avoid it.

Hailuo 3 Max takes no reference images. Wire a still into the node's reference input and it is used as the opening frame instead, not carried as a subject into a new scene. For a subject that has to appear somewhere it was never photographed, use a model with reference input, such as HeyGen Video or Wan 2.7.

Resolution: 480p for drafts, 768p for delivery

768p is both the ceiling and the default. Leave Resolution on Model default and a 9:16 clip comes back at 768 by 1344. Pick 480p and it comes back at 480 by 832, for five eighths of the price.

Neither is 1080p. For phone-first vertical work that rarely matters. For a clip shown large, run it through the Video Upscale node, render that shot on Hailuo 3 at 2K or 4K, or pick a model that goes higher, such as Veo 3.1 Lite at up to 1080p.

About the audio

The catalog lists no audio setting for this endpoint, and the clips we rendered carried only a quiet ambient track, well below normal listening level, even on a prompt that asked for natural ambient sound. Treat Hailuo 3 Max as picture.

If the clip needs sound, add it in the graph. A Text to Speech node handles narration, an Audio Generation node handles music or effects, and the Sequence node mixes them over the picture. If the soundtrack has to come from the model itself, Hailuo 3 generates audio with every clip.

Two lanes to the same model

Treza reaches H3 Max two ways, and the Video Generation node's picker lists both. The OpenRouter lane is the one this post describes: keyframes at both ends, 5 to 15 seconds, six aspect ratios. The second lane, labeled MiniMax H3 Max (fal), is text-to-video only, with no keyframes, 5 to 10 second clips, and the same 5 and 8 cent rates; fal quotes a 5 second clip in under 3 seconds on it. Pick the fal lane for the fastest prompt-only draft, and the OpenRouter lane the moment a shot needs a frame pinned.

What you actually pay on Treza

Hailuo 3 Max through a Treza pipeline is metered from a prepaid credit balance. One credit equals $0.01, credit packs start at $5, credits never expire, and a failed generation charges nothing.

Before a run starts, the platform estimates the media cost of the graph and refuses to start one your balance cannot cover. For Hailuo 3 Max the estimator holds the listed rate for the resolution you pick: 8 cents of provider cost per second at 768p or with no resolution set, and 5 cents at 480p. With no pricier tier above 768p and no input that raises the rate, the hold matches the charge. Hailuo 3, by contrast, is held at its 4K rate of 16 cents a second whatever the node names.

For a finished video costed end to end, with narration and captions in the graph rather than just the render, see the AI video cost breakdown.

Calling Hailuo 3 Max as an API without building the integration

Direct API access to a video model means credentials, a submit-and-poll job pattern, retries, moderation rejections, and somewhere to store the output.

On a Treza AI video pipeline, Hailuo 3 Max is a node on a canvas. Set the duration, aspect ratio, and resolution, wire a first frame or both frames into it, place it next to a script, voiceover, or captioning step, and publish the whole chain as one versioned endpoint. Your application calls a typed /invoke route with JSON in and JSON out, or points an existing OpenAI SDK at the OpenAI-compatible route.

The swap is what pays off later. When the next model wins on price or quality, you change the dropdown and republish, and the endpoint your product calls does not change. Our hosted MCP server exposes the same pipelines as tools, so an agent in Claude or ChatGPT can render a brief and return the link.

Where Hailuo 3 Max earns its place in a graph

Three jobs where the keyframes do the work. Product shots that start on the photo, where the photo is wired in as the first frame and the camera moves off it. Transitions between two stills, a sketch to a render or a before to an after, where the last frame pins the landing. And short vertical cuts for Shorts and Reels, drafted at 480p for 25 cents per 5 seconds and rendered at 768p once the framing is right.

The fastest way to learn what it costs for your own work is to run your real prompt at your real duration and read the charge. Open the AI video generator, pick Hailuo 3 Max, and generate once.

Frequently Asked Questions

How much does Hailuo 3 Max cost per second?

$0.05 per second at 480p and $0.08 per second at 768p, the listed rates for the endpoint we run as of October 2026, and our test renders billed exactly that. A first or last frame does not change the rate. On Treza the charge comes from a prepaid credit balance, and a failed generation charges nothing.

How is Hailuo 3 Max different from Hailuo 3?

Hailuo 3 Max is a further-trained version of MiniMax H3. On the endpoints we run, Max takes a first frame or a first and last frame, renders at 480p or 768p, and has no audio setting; Hailuo 3 takes no keyframes, renders up to 4K, and generates audio with every clip. At 768p Max costs two cents a second more.

Does Hailuo 3 Max take a last frame?

Yes. The Video Generation node takes a first frame alone, or a first and a last frame together, and the model fills in the motion between them. Reference images are not supported on this endpoint; a still wired in as a reference is used as the opening frame instead.

What resolution does Hailuo 3 Max render?

480p or 768p. With no resolution set it renders 768p, which is 768 by 1344 at 9:16. It does not render 1080p, 2K, or 4K, so for a clip shown large, upscale it with the Video Upscale node or render that shot on Hailuo 3.

Does Hailuo 3 Max generate audio?

There is no audio setting, and the clips we rendered carried only a quiet ambient track. Plan sound in the graph with Text to Speech or Audio Generation, mixed in the Sequence node, or render on Hailuo 3 when the soundtrack has to come from the model.

Can I call Hailuo 3 Max through an API without a MiniMax account?

Yes. On a Treza pipeline, Hailuo 3 Max is a node you configure once and publish as a versioned HTTP endpoint, with a typed /invoke route and an OpenAI-compatible route. Generations are metered from prepaid credits that start at $5 and never expire.