MiniMax Hailuo 03 Max: playground and API

Model guide

Build with MiniMax Hailuo 03 Max.

MiniMax Hailuo 03 Max: higher-fidelity video from text or a first frame, at 480P or 768P

Best for

MiniMax Hailuo 03 Max: higher-fidelity video from text or a first frame, at 480P or 768P Use the playground to set up a single step, then open a workflow when you want to connect more tools.

Inputs explained

Prompt (required): Set the prompt used by this model. Type: text.
Image (optional): Set the image used by this model. Type: image.
End Frame (optional): Set the end frame used by this model. Type: image.
Reference Images (max 4, aspect 0.4-2.5) (optional): Set the reference images (max 4, aspect 0.4-2.5) used by this model. Type: list of image.
Reference Videos (max 3, 2-15s each, 15s total) (optional): Set the reference videos (max 3, 2-15s each, 15s total) used by this model. Type: list of video.
Reference Audio (max 3, 2-15s each, 15s total) (optional): Set the reference audio (max 3, 2-15s each, 15s total) used by this model. Type: list of audio.
Resolution (optional): Set the resolution used by this model. Type: text. Options: 480P, 768P, 1080P. Default: "768P".
Aspect Ratio (optional): Set the aspect ratio used by this model. Type: text. Options: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16. Default: "16:9".
Duration (seconds) (optional): Set the duration (seconds) used by this model. Type: number. Minimum: 5. Maximum: 15. Default: 5.
Prompt Expansion (required): Set the prompt expansion used by this model. Type: text. Options: disabled, balanced, quality. Default: "balanced".

Tips before you run

Start with the default settings and change one input at a time. Keep the result you like on your board so the next step can use it.

If you add a reference file, use a link the workflow can read. Check the inputs and credit estimate on the board before starting a run.

Working with this model

One node, two endpoints

MiniMax ships Hailuo 03 Max as two separate APIs. This is one node, and what you wire decides which API the run goes to:

nothing but a Prompt goes to text-to-video
an Image goes to image-to-video, and that frame becomes the first frame

There is no reference-to-video on Max. If you need reference images or reference audio, use the plain MiniMax Hailuo 03 node instead.

Resolution is the price

Resolution is the one setting that changes what a run costs. 768P is the default and roughly 60% dearer per second than 480P. Cost is per second of output, so resolution and Duration multiply: a 15 second 768P render is six times a 5 second 480P one. Duration is a whole number of seconds, 5 to 15.

Hailuo 03 Max tops out at 768P. If you want 2K, that is the plain Hailuo 03 node, which renders at 2K only and costs considerably more per second.

Aspect ratio only applies to text-to-video

Aspect Ratio is read on the text-to-video path. It is ignored for image-to-video, where the output follows the shape of the image you wired.

End Frame needs a start frame

End Frame gives you first-to-last keyframe control, and it is only read on the image-to-video path. Wiring an end frame on its own does nothing, because reaching that path requires an Image. Wire both.

Prompt Expansion

Prompt Expansion decides how much MiniMax rewrites your prompt before generating. "balanced" is the default. "disabled" sends your text through untouched, which is what you want when you have already written a detailed shot description and do not want it embellished. "quality" leans hardest on the rewrite and suits a short prompt.

Prompts can run to 50,000 characters, so there is no practical reason to trim a detailed one.

Common questions

How do I use MiniMax Hailuo 03 Max on Wireflow?

Set the inputs above and choose Open in a workflow. Sign in to save your board, then check its credit estimate before you run.

Can I use MiniMax Hailuo 03 Max through an API?

Yes. Save a workflow with this model, then call the Wireflow run API with its workflow ID. The API tab has examples you can copy.

What changes the credit price?

The price shown uses the model defaults. Your inputs and model settings can change the total. The workflow shows an estimate before you run.

Can I connect this to another model?

Yes. Open it in a workflow and connect its output to another node with a matching input type.

Input

Result Idle