MiniMax H3 API

Access MiniMax H3 (Hailuo 03 / Hailuo 3) through EvoLink's unified video API. Test 2K text-to-video, first/last-frame image-to-video, and reference-to-video.

MiniMaxVideo GenerationAvailable

From ~$0.076 / s for the 768p video rate$0.080 official price-5%

Text-to-VideoImage-to-VideoReference-to-VideoFirst / Last FrameMultimodal ReferenceAsync API

Production routeLive

Live

Model Highlights

MiniMax's next-generation Hailuo video route at 2K or 768p

Use Cases

Ad creative, ecommerce, narrative shorts, multi-model pipelines

Input

Text, first/last frame, or reference media

Output

2K / 768p video (MP4), 4-15s

Playground Pricing Readme API

Try MiniMax H3 before you integrate the API.

Test MiniMax H3 output, estimate credits, and pick the model ID that matches your input type before wiring the async API.

Create a Sample

Pick a route, enter your prompt, and set the parameters.

Prompt

Browse prompt examples

Create a 5-second horizontal 16:9 cinematic short film.

Keep the woman highly consistent: young woman with dark wavy hair and bangs, black shirt, calm but tense expression, sitting alone in the back seat of a car. Keep the rainy car window, red and cyan neon reflections, dark interior, realistic city night atmosphere, shallow depth of field, and intimate film-noir mood.

Story conflict: she is not simply riding in a car. She has just made a decision that will betray someone she once trusted. Outside, someone is following her through the rain. She must choose whether to keep going or disappear before the car reaches the meeting point.

0-1s: Rain slides slowly down the car window. She sits still in the back seat, looking through the glass. Red neon light washes across her face like a warning.

1-2s: A phone on her lap vibrates and lights up, but the screen shows no readable text. She glances down, does not answer, and quietly tightens her hand around it.

2-3s: A bright pair of headlights appears behind the car, reflected in the wet window. Her eyes shift toward the reflection. She realizes the same car is still following.

3-4s: She stays calm. No panic. She turns the phone face down on the seat, then reaches into her coat and touches an old photograph or small envelope, as if saying goodbye to her old life.

4-5s: She looks toward the driver without raising her voice. Her lips barely move, silently implying: "Don't stop." Red light flashes across the window. End on her close-up through the rain-covered glass - no tears, no fear, only a cold final decision.

Style: realistic cinematic thriller, urban rainy night, neon reflections, film noir mood, restrained acting, subtle emotional tension, natural face movement, real skin texture, wet glass detail, low-light cinematography, soft film grain, shallow depth of field, intimate close-up, high-end independent crime film opening.

Avoid: exaggerated car chase, explosions, guns, fighting, crying, screaming, fake acting, overdramatic gestures, plastic skin, AI-looking face, changing identity, changing outfit, random text, subtitles, watermark, logo, distorted face, chaotic camera movement.

Duration5s

Quality

Aspect RatioAuto isn't available for Text-to-Video — pick an explicit aspect ratio.

Estimated: ~$0.61841.99 cr

2K × 5s

Ready

History

Results are kept for 24 hours — download anything you want to keep.

Your generation history will appear here

MiniMax H3 pricing is per output second.

One per-second rate covers all three routes; 768p is billed at 8/13 of the 2K rate. Reference video seconds are added to the output duration, so a reference-guided clip costs more than a text prompt of the same length.

First Test Cost

Text-to-Video, 4s at 768P — the cheapest possible request, no input media

Minimum charge

Credits20.672

One 4s 768P sample

Approx. Cost$0.304

About 32 tests with $10 in credits.

This is also the minimum charge: 4s at 768p is the cheapest valid request, so no MiniMax H3 task costs less.

EvoLink rate $0.304

official price $0.320 · You save 5%

Testing Budget Guide

At the first-test settings: 4s at 768P

Add Credits

$10

About 32 tests

Good for a first validation.

$50

About 164 tests

Good for prompt and reference iteration.

$100

About 328 tests

Good before production integration.

Common Cost Examples

Short text-to-video testT2V 4s 768p

~$0.304-5%

~20.672 cr·~$0.320 official price

Full-length clipT2V 15s 768p

~$1.140-5%

~77.52 cr·~$1.200 official price

Reference video testR2V 4s + 3s input 768p

~$0.532-5%

~36.176 cr·~$0.560 official price

Estimates use the current per-second rate for your user group. Reference video seconds are billed with the output duration.

MiniMax H3 Pricing

Mode	768P	2K	Billing
Text-to-VideoStandard route	$0.076/s-5% 5.168 cr/s$0.080 official price	$0.124/s-5% 8.398 cr/s$0.130 official price	Output video seconds
Image-to-VideoStandard route	$0.076/s-5% 5.168 cr/s$0.080 official price	$0.124/s-5% 8.398 cr/s$0.130 official price	Output video seconds
Reference-to-VideoImage or audio reference	$0.076/s-5% 5.168 cr/s$0.080 official price	$0.124/s-5% 8.398 cr/s$0.130 official price	Output video seconds
Reference-to-VideoVideo reference	$0.076/s-5% 5.168 cr/s$0.080 official price	$0.124/s-5% 8.398 cr/s$0.130 official price	Output + reference video seconds

Text-to-Video

Standard route · Output video seconds

768P $0.076/s · 5.168 cr/s $0.080/s-5%

2K $0.124/s · 8.398 cr/s $0.130/s-5%

Image-to-Video

Standard route · Output video seconds

768P $0.076/s · 5.168 cr/s $0.080/s-5%

2K $0.124/s · 8.398 cr/s $0.130/s-5%

Reference-to-Video

Image or audio reference · Output video seconds

768P $0.076/s · 5.168 cr/s $0.080/s-5%

2K $0.124/s · 8.398 cr/s $0.130/s-5%

Reference-to-Video

Video reference · Output + reference video seconds

768P $0.076/s · 5.168 cr/s $0.080/s-5%

2K $0.124/s · 8.398 cr/s $0.130/s-5%

Billed per second of output video. Reference images and reference audio do not add input seconds; reference videos do.

Minimum charge per task equals the cheapest valid request (768p x 4s). Failed, expired, and rejected tasks are fully refunded.

Item	Use Case	Price	Credits
Input imagesFrom image #6	Applies when a request sends more than 5 input images.	$0.038 / image-5% $0.040 official price	2.584 cr

The first 5 input images of a request are included; billing starts from image #6. Each additional image is charged per image on top of the per-second cost, and the minimum charge applies to the per-second part only.

What MiniMax H3 is best for.

MiniMax H3 is MiniMax's next-generation multimodal model in the Hailuo AI video family, not the MiniMax M3 language model for coding and agents. It runs on EvoLink as three routes — text, frame, and reference driven — behind one API key and one credit balance.

Open the playground

Input

Text / First-last frame / Reference media

Output

MP4 video

Duration

4-15 seconds

Resolution

768p / 2K

References

Up to 12 files: 9 images, 3 videos, 3 audio

MiniMax H3: three routes, three model IDs

Each generation mode is its own model ID behind the same API key and credit balance - pick the route that matches your input.

MiniMax H3 Text-to-Video

minimax-h3-text-to-video

Generate 2K or 768p video from a text prompt.

MiniMax H3 Image-to-Video

minimax-h3-image-to-video

Animate a first frame, a last frame, or both.

MiniMax H3 Reference-to-Video

minimax-h3-reference-to-video

Generate video guided by image, video, and audio references.

MiniMax H3 model information

The verified request contract as exposed by EvoLink. Anything MiniMax has not published is left out rather than guessed.

Model type

Video generation

MiniMax H3 belongs to the Hailuo video family, not the MiniMax M3 language-model family.

Supported inputs

Text, first/last frame, reference image + video + audio

Frame inputs and reference inputs are mutually exclusive upstream, which is why EvoLink exposes three separate model IDs.

API model IDs

minimax-h3-text-to-video / -image-to-video / -reference-to-video

Pick the ID that matches your input type; the request body is validated against a strict whitelist per route.

Output and duration

2K / 768p, 4-15s, MP4

768p is open and is the default when quality is omitted; it is billed at 8/13 of the 2K rate.

API pricing

Per output second, 768p and 2K tiers

Reference video seconds are added to the billable duration; reference images and audio are not.

Task lifecycle

Async submit, poll or callback, no cancel

Tasks report pending / processing / completed / failed. Cancellation is not offered upstream, so failures refund instead.

Key MiniMax H3 API parameters

These fields cover the main integration decisions. Open the API Reference for route-specific request and response examples.

modelstring

Choose the text, image, or reference EvoLink model ID that matches your input.

Default: route-specific

promptstring

Describe the scene, motion, and camera behavior in up to 7000 characters.

Default: required

durationinteger

Set 4 through 15 output seconds; billing follows the output duration.

Default: 4

qualityenum

Select 768p or 2k. 768p is billed at 8/13 of the 2K per-second rate.

Default: 768p

aspect_ratioenum

Text-to-Video requires an explicit ratio (21:9, 16:9, 4:3, 1:1, 3:4, 9:16); Reference-to-Video defaults to adaptive; Image-to-Video follows the input frames.

Default: 16:9 / adaptive

image_start / image_endstring

Image-to-Video only: HTTP(S) URLs for a first frame, a last frame, or both.

Default: optional

image_urls / video_urls / audio_urlsstring[]

Reference-to-Video only: up to 9 images, 3 videos, and 3 audio tracks (12 files in total); reference video seconds are billed.

Default: []

callback_urlstring

Public HTTPS URL that receives the final task state; keep the task ID as a polling fallback.

Default: optional

Two ways to use MiniMax H3: EvoLink API or Agent

Use the EvoLink API for product integration and batch jobs, or call it from Codex, Claude, or Gemini for fast creative and development workflows. Both paths share the same EvoLink API key, balance, model routes, and task history.

Option 1

Integrate with the EvoLink API

Best for: product backends, batch jobs, automated workflows

Call EvoLink’s unified video API from your server and control the model ID, parameters, task queue, callbacks, and result storage.

1Validate output and cost with a real brief in Playground
2Create an EvoLink API key in the console
3Pick the model ID that matches your task from the route cards above
4Submit the task and retrieve the result by polling or HTTPS callback

Create EvoLink API Key

Option 2

Call it with an Agent

Best for: creative and development tasks in Codex, Claude, and Gemini

Give the Agent your output goal, assets, and acceptance criteria. It can choose the route, assemble the request, track the task, and return the result without requiring you to hand-code every step.

1Set EVOLINK_API_KEY in your local environment; never put it in code or a prompt
2Describe the output goal, references, and the key parameters
3Ask the Agent to call a MiniMax H3 route and save the task ID
4Let the Agent poll the final state, download the result, and report failure reasons

Where MiniMax H3 fits in a video workflow

The same workflows the Hailuo family is evaluated on, now callable through a verified route.

Text-to-video creation with MiniMax H3

Prompt adherence, camera direction, motion, and physical plausibility for storyboards, social clips, and creative concepts.

Image-to-video animation with MiniMax H3

Animate a first frame, land on a last frame, or drive a transition between both while keeping product shape and character identity.

Advertising creative with MiniMax H3

Batch throughput, retries, brand consistency, and cost per accepted ad rather than one showcase clip.

Ecommerce product video with MiniMax H3

Product fidelity, controllable movement, aspect-ratio delivery, and logo handling before customer-facing use.

Narrative and short-form series with MiniMax H3

Character consistency and shot continuity across a sequence, with reference images and reference video to anchor style.

Multi-model video products with MiniMax H3

Run MiniMax H3 as one route beside Seedance, Kling, and Wan, and keep a tested fallback instead of coupling to one provider.

Try these workflows in Hailuo Video Generator

MiniMax H3 API code example and error handling

This example shows the shortest runnable flow: create a task, save the returned task ID, then poll or wait for an HTTPS callback. Open the API tab for the complete parameter and response reference.

View complete API docs

cURL

curl -X POST https://api.evolink.ai/v1/videos/generations \
  -H "Authorization: Bearer $EVOLINK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "minimax-h3-text-to-video",
    "prompt": "A female captain before a vast observation window as the fleet jumps away, cinematic slow dolly-in",
    "duration": 5,
    "quality": "768p",
    "aspect_ratio": "16:9",
    "callback_url": "https://your-domain.com/webhooks/video-task-completed"
  }'

# Save the returned task ID, then query:
curl https://api.evolink.ai/v1/tasks/{task_id} \
  -H "Authorization: Bearer $EVOLINK_API_KEY"

Parameter validation failed

Check every value against the parameter cards above: ranges, allowed options, asset counts, and file types must match the selected route.

Authentication or balance issue

Check the Authorization bearer token and confirm the available balance and task billing in the console.

Content or asset rejected

Review likeness rights, brand assets, readable text, sensitive content, and input-asset compliance.

Task failed or timed out

A MiniMax H3 task that reaches a final failed state is not charged. Keep the task ID, inspect the final state, and identify whether the cause is validation, review, or execution before retrying.

Callback not received

callback_url must be a public HTTPS URL, not localhost or a private network address. Always keep the task ID as a polling fallback.

Is MiniMax H3 worth adding to your video stack?

Choose it for the request contract you can verify today, and keep another route ready when the workload exceeds those limits.

Good fit when

You need 2K video clips from 4 to 15 seconds.
Your product needs text-to-video, first/last-frame image-to-video, and reference-to-video.
Reference images, videos, or audio are part of the creative brief.
Your application already supports async polling or callbacks.
You evaluate accepted-output cost across multiple video models.

Consider another route when

You require 4K output or clips longer than 15 seconds.
The task must return synchronously or remain cancellable after submission.
You need a silent clip — H3 puts audio on every generation.
You cannot save the completed MP4 within the 24-hour result window.
You need open weights or local deployment.

Run a 4-second test

How to decide whether MiniMax H3 fits production

Use real workloads and acceptance criteria — not selected showcase clips — to measure usable output, consistency, total cost, and operating reliability.

Does it produce more usable clips?

Measure prompt and reference adherence, temporal consistency, visible artifacts, and accepted outputs across the full attempt set.

Does it preserve identity and products?

Compare character, face, clothing, logo, product shape, composition, and style retention from the same source assets.

Does it improve the full workflow?

Include queue limits, failed-task refunds, moderation, retries, reviewer correction, delivery format, and downstream editing — not only generation speed.

Is it worth adding to production?

Calculate cost per final accepted clip, and confirm callbacks, retention rules, monitoring, rollback, and multi-model fallback fit your product.

How MiniMax H3 fits your video stack

Feature	MiniMax H3	Hailuo 2.3	Other video routes
Primary role	Next-generation Hailuo route	Documented Hailuo baseline	Cross-vendor comparison and fallback
Generation modes	T2V + I2V (first/last frame) + R2V	T2V + I2V	Model-dependent
Reference input	Image + video + audio	Image only	Model-dependent
Resolution	2K / 768p	768p / 1080p	480p to 4K
Duration	4-15s	6s / 10s	3-15s
Billing	Per output second (+ reference video seconds)	Per video	Per second or per video

Primary role

MiniMax H3Next-generation Hailuo route

Hailuo 2.3Documented Hailuo baseline

Other video routesCross-vendor comparison and fallback

Generation modes

MiniMax H3T2V + I2V (first/last frame) + R2V

Hailuo 2.3T2V + I2V

MiniMax H3 key details

Three MiniMax H3 Model IDs

Use minimax-h3-text-to-video, minimax-h3-image-to-video, or minimax-h3-reference-to-video depending on whether you send a prompt, frames, or reference media.

MiniMax H3 strict parameter whitelist

Only the documented fields are accepted. Legacy aliases such as seed, image_url, or first_frame_image return 400 and name the offending field instead of being silently ignored.

MiniMax H3 async production flow

Submit a task, poll status or use a callback, then fetch the finished MP4. Cancellation is not available; failed, expired, and rejected tasks are refunded in full.

Why use the MiniMax H3 API through EvoLink

Cost Clarity

See credits and per-second cost, including reference video seconds, before each generation.

Unified API

Switch between Hailuo, Seedance, Kling, and other video routes without rebuilding auth or billing.

Unified Balance

Track credits, spending, and model usage for every route in one console.

Task Status Tracking

See whether a task is pending, processing, completed, or failed, with refunds on failure.

Model Choice

Keep MiniMax H3 as one tested route and fall back to another family when a workload needs it.

Hailuo Model Family

Compare Hailuo generations under one EvoLink account and API key, then choose by resolution, duration, reference workflow, and cost.

Compare all Hailuo models

MiniMax H3 Max

The fast H3 route for 480p / 768p text-to-video and first- or last-frame generation from 5 to 15 seconds.

View

Hailuo 2.3

The documented Hailuo generation for text-to-video and image-to-video through Standard and Fast routes.

View

Hailuo 02

Full-featured earlier generation with T2V, I2V, and first-last-frame modes at 512p / 768p / 1080p.

View

Compare MiniMax H3 before you adopt it

Use the family page and the dated comparisons for the decision that brought you here.

Read the MiniMax H3 API guide

cURL, Python, and TypeScript examples for the three routes, async submission, polling, callbacks, and error handling.

MiniMax H3 release timeline

Confirmed evidence, reported timing, and the verification checklist EvoLink used before opening the route.

MiniMax H3 vs Hailuo 2.3

Decide when the next generation has enough evidence to justify an upgrade inside the same family.

MiniMax H3 vs Seedance 2.0

Compare MiniMax H3 with an available multimodal audio-video workflow.

MiniMax H3 vs Seedance 2.5

Decide whether to ship on H3 or move to the Seedance 2.5 API, live on EvoLink since August 7, 2026.

MiniMax H3 FAQ

What is MiniMax H3?

MiniMax H3 is MiniMax's next-generation multimodal model in the Hailuo AI video family. It is a video model, not MiniMax M3 for coding and agent workloads.

Are the MiniMax H3 API, Hailuo 03 API, and Hailuo 3 API the same?

Yes. The MiniMax H3 API, Hailuo 03 API, and Hailuo 3 API refer to the same three Hailuo-family video routes on EvoLink.

What can MiniMax H3 be used for?

Text-to-video, image-to-video with first and last frame control, reference-guided generation, advertising creative, ecommerce product video, narrative shorts, and multi-model video products.

Is MiniMax H3 available through EvoLink?

Yes. EvoLink exposes MiniMax H3 as three routes through the unified video endpoint, with API keys, credit balance, async task tracking, and refunds on failed tasks.

What are the MiniMax H3 API model IDs?

minimax-h3-text-to-video, minimax-h3-image-to-video, and minimax-h3-reference-to-video. Pick the ID that matches your input type; each route validates its own parameter set.

How much does the MiniMax H3 API cost?

MiniMax H3 is billed per second of output video. 768p is the default tier and is billed at 8/13 of the 2K rate. Use the pricing section and playground estimator for the current per-second rate of your user group before you generate.

How is reference video billed?

Reference video seconds are added to the output duration, so a 5-second clip with a 3-second reference video is billed as 8 seconds. Reference images and reference audio do not add input seconds.

What input limits apply to MiniMax H3 reference media?

Reference-to-video accepts up to 9 images, 3 videos, and 3 audio files. Videos and audio must be 2 to 15 seconds each and no more than 15 seconds total per media type. Audio can be used on its own as the only reference input.

Does MiniMax H3 support 4K, longer clips, or a cancel endpoint?

No. Output is 2K or 768p with durations of 4 to 15 seconds, and there is no upstream cancel endpoint, so tasks report can_cancel false and failures are refunded in full.

How does MiniMax H3 compare with Hailuo 2.3?

Hailuo 2.3 remains the per-video Standard and Fast baseline at 768p / 1080p. MiniMax H3 adds reference video and audio inputs, first-last-frame control, 2K / 768p output, and per-second billing.

How do I get the completed MiniMax H3 video?

Submit the async task, poll GET /v1/tasks/{task_id} or use a callback, then download the MP4 from the completed task response.

How long is a MiniMax H3 result URL available?

The completed video URL is available for 24 hours. Copy the MP4 into your own authorized storage before the link expires.

MiniMax H3 API

Try MiniMax H3 before you integrate the API.

MiniMax H3 pricing is per output second.

Common Cost Examples

MiniMax H3 Pricing

What MiniMax H3 is best for.

MiniMax H3: three routes, three model IDs

MiniMax H3 Text-to-Video

MiniMax H3 Image-to-Video

MiniMax H3 Reference-to-Video

MiniMax H3 model information

Model type

Supported inputs

API model IDs

Output and duration

API pricing

Task lifecycle

Key MiniMax H3 API parameters

Two ways to use MiniMax H3: EvoLink API or Agent

Integrate with the EvoLink API

Call it with an Agent

Where MiniMax H3 fits in a video workflow

Text-to-video creation with MiniMax H3

Image-to-video animation with MiniMax H3

Advertising creative with MiniMax H3

Ecommerce product video with MiniMax H3

Narrative and short-form series with MiniMax H3

Multi-model video products with MiniMax H3

MiniMax H3 API code example and error handling

Parameter validation failed

Authentication or balance issue

Content or asset rejected

Task failed or timed out

Callback not received

Is MiniMax H3 worth adding to your video stack?

Good fit when

Consider another route when

How to decide whether MiniMax H3 fits production

Does it produce more usable clips?

Does it preserve identity and products?

Does it improve the full workflow?

Is it worth adding to production?

How MiniMax H3 fits your video stack

MiniMax H3 key details

Three MiniMax H3 Model IDs

MiniMax H3 strict parameter whitelist

MiniMax H3 async production flow

Why use the MiniMax H3 API through EvoLink

Cost Clarity

Unified API

Unified Balance

Task Status Tracking

Model Choice

Hailuo Model Family

Other video models on EvoLink besides MiniMax H3

Compare MiniMax H3 before you adopt it

Read the MiniMax H3 API guide

MiniMax H3 release timeline

MiniMax H3 vs Hailuo 2.3

MiniMax H3 vs Seedance 2.0

MiniMax H3 vs Seedance 2.5

MiniMax H3 FAQ

What is MiniMax H3?

Are the MiniMax H3 API, Hailuo 03 API, and Hailuo 3 API the same?

What can MiniMax H3 be used for?

Is MiniMax H3 available through EvoLink?

What are the MiniMax H3 API model IDs?

How much does the MiniMax H3 API cost?

How is reference video billed?

What input limits apply to MiniMax H3 reference media?

Does MiniMax H3 support 4K, longer clips, or a cancel endpoint?

How does MiniMax H3 compare with Hailuo 2.3?

How do I get the completed MiniMax H3 video?

How long is a MiniMax H3 result URL available?