MiniMax video model on Crazyrouter
All Hailuo (MiniMax) models →

Generate video with the Hailuo H3 API

Hailuo H3 runs on Crazyrouter behind a unified asynchronous video endpoint: submit a task, poll for the result, pay per generated second.

Video generationText to videoImage to video
Get API KeyView live pricing

Paste the model ID, live price, base URL and a working request straight into your coding agent.

Hailuo (MiniMax)
hailuo-h3
$0.071–$0.251
Price range
Model ID
hailuo-h3
Provider
MiniMax
Endpoint
/minimax/v1/video_generation
Task
video generation
Input modalities
Text, image
Output modalities
Video

What is Hailuo H3?

Hailuo H3 is a Hailuo video model from MiniMax for text-to-video and image-to-video generation. Crazyrouter wraps it in the same task API used for every video model in the catalogue, so switching between vendors is a one-line change.

Unified task API

One submit/poll flow for every video model on Crazyrouter.

Per-second billing

Charged for generated seconds at the resolution-specific rate; failed tasks are not billed.

Text and image inputs

Start from a prompt or a reference frame where the upstream model supports it.

Shared balance

Same key and balance as your chat and image workloads.

Hailuo H3 API quickstart

Submit a job to https://api.crazyrouter.com/minimax/v1/video_generation with model hailuo-h3, then poll the returned task id until the status is completed.

cURL
curl https://api.crazyrouter.com/minimax/v1/video_generation \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $CRAZYROUTER_API_KEY" \
  -d '{
    "model": "hailuo-h3",
    "prompt": "Slow aerial drone shot gliding over a small coastal town at sunrise, pastel houses, calm sea, soft golden light, cinematic, steady camera",
    "duration": 6,
    "resolution": "768P"
  }'
Show OpenAI-compatible migration example
cURL
curl "https://api.crazyrouter.com/minimax/v1/query/video_generation?task_id=$TASK_ID" \
  -H "Authorization: Bearer $CRAZYROUTER_API_KEY"

Use this model via API

This model is available through the paid API. Create an account, add funds and create an API key. The $2 Playground credit is for supported text models only, not API calls.

Sign up for API access

Hailuo H3 sample outputs

Every sample below was generated through Crazyrouter with the request shown above, unretouched and uncurated; latency and cost are what that single run actually took.

Coastal sunrise, droneThis run $0.429
Show full prompt

Slow aerial drone shot gliding over a small coastal town at sunrise, pastel houses, calm sea, soft golden light, cinematic, steady camera

Product turntableLatency 132s · This run $0.429
Show full prompt

A white running shoe rotating slowly on a turntable in a bright studio, seamless light grey background, soft shadows, product commercial

Tracking shot, crosswalkLatency 135s · This run $0.429
Show full prompt

A person in a yellow raincoat walking across a busy crosswalk in Tokyo at dusk, tracking shot from behind, realistic motion, shallow depth of field

2D anime forestLatency 137s · This run $0.429
Show full prompt

2D anime style: a forest path in autumn with falling leaves, gentle wind, warm afternoon light, hand-painted background, smooth animation

Generated 2026-09-28

Best use cases

Short-form social video from a prompt
Product demos and explainers
Storyboard and previz iterations
Animating stills for ads
Batch generation pipelines
Prototyping video features in apps

Works with developer tools

HTTP clients

Plain REST: POST to submit, GET to poll.

n8n / Dify workflows

Use HTTP nodes with a wait/poll loop.

Custom backends

Store the task id and poll from a job queue.

Crazyrouter console

Track task status and cost per job.

Hailuo H3 FAQ

Is generation synchronous?

No. The submit call returns a task id; poll the matching query endpoint (second example above) until it completes.

How is it billed?

Per generated second at the rate for the requested resolution and mode; the range on this page covers the available options.

Which durations and resolutions are available?

Those exposed by the upstream model; pass duration, resolution and aspect ratio as documented.

Can I start from an image?

Yes for image-to-video capable models: include the image URL in the request.

How long do results stay available?

Download the result promptly after completion; upstream links expire.