MiniMax H3 API

video

Video generation by minimax

MiniMax H3 video generation at 2K resolution with keyframe and reference image control.

Call MiniMax H3 from Node.js, Python, or any HTTP client. No GPU provisioning, no cold starts, no Docker images — just an API key and a prompt.

Get your API keySee pricing

Free credits included. Powering 1M+ generations daily.

Start in 60 seconds

Install the official SDK for Node.js or Python, drop in your API key, and generate your first video.

npm install getimg-ai
import GetimgAI from "getimg-ai";

const client = new GetimgAI();

const submission = await client.videos.generations.create({
    model: "minimax-h3",
    prompt: "Create a restless backstage sequence immediately before a live performance.",
    aspect_ratio: "21:9",
    resolution: "2K",
    duration: 5
  });
const requestId = submission.id;

// Poll
let result;
while (true) {
  result = await client.videos.generations.retrieve(requestId);

  if (result.status === "completed") break;
  if (result.status === "failed") {
    throw new Error(result.error.message);
  }

  await new Promise((r) => setTimeout(r, 5000));
}

console.log(result.data[0].url);

How MiniMax H3 API works

Submit a generation request, poll for completion, download the video.

1

Get your API key

Create a key in the Developer Dashboard. Free credits included.

2

Send a request

POST to /v2/videos/generations with your prompt, duration, and resolution.

3

Get your video

Poll the status endpoint. When it completes, download the video from the returned URL.

What you can build with MiniMax H3

Every example uses the same authentication and works with the official getimg-ai and getimg SDKs, or plain HTTP.

Text to video

Generate a video using just a text prompt.

import GetimgAI from "getimg-ai";

const client = new GetimgAI();

const submission = await client.videos.generations.create({
    model: "minimax-h3",
    prompt: "Create a restless backstage sequence immediately before a live performance.",
    aspect_ratio: "21:9",
    resolution: "2K",
    duration: 5
  });
const requestId = submission.id;

// Poll
let result;
while (true) {
  result = await client.videos.generations.retrieve(requestId);

  if (result.status === "completed") break;
  if (result.status === "failed") {
    throw new Error(result.error.message);
  }

  await new Promise((r) => setTimeout(r, 5000));
}

console.log(result.data[0].url);

Image to video

Animate a still image as the starting frame.

import GetimgAI from "getimg-ai";

const client = new GetimgAI();

const submission = await client.videos.generations.create({
    model: "minimax-h3",
    prompt: "Animate the scene with subtle camera movement and natural motion.",
    aspect_ratio: "21:9",
    resolution: "2K",
    duration: 5,
    images: [
      {
        url: "https://your-bucket.s3.amazonaws.com/first.png",
        role: "first_frame"
      }
    ]
  });
const requestId = submission.id;

// Poll
let result;
while (true) {
  result = await client.videos.generations.retrieve(requestId);

  if (result.status === "completed") break;
  if (result.status === "failed") {
    throw new Error(result.error.message);
  }

  await new Promise((r) => setTimeout(r, 5000));
}

console.log(result.data[0].url);

First & last frame

Interpolate motion between a starting and ending frame.

import GetimgAI from "getimg-ai";

const client = new GetimgAI();

const submission = await client.videos.generations.create({
    model: "minimax-h3",
    prompt: "Smooth transition from the first to the last frame.",
    aspect_ratio: "21:9",
    resolution: "2K",
    duration: 5,
    images: [
      {
        url: "https://your-bucket.s3.amazonaws.com/first.png",
        role: "first_frame"
      },
      {
        url: "https://your-bucket.s3.amazonaws.com/last.png",
        role: "last_frame"
      }
    ]
  });
const requestId = submission.id;

// Poll
let result;
while (true) {
  result = await client.videos.generations.retrieve(requestId);

  if (result.status === "completed") break;
  if (result.status === "failed") {
    throw new Error(result.error.message);
  }

  await new Promise((r) => setTimeout(r, 5000));
}

console.log(result.data[0].url);

Reference image

Guide the generation with a character, style, or content reference.

import GetimgAI from "getimg-ai";

const client = new GetimgAI();

const submission = await client.videos.generations.create({
    model: "minimax-h3",
    prompt: "A cinematic shot inspired by the provided reference image.",
    aspect_ratio: "21:9",
    resolution: "2K",
    duration: 5,
    images: [
      {
        url: "https://your-bucket.s3.amazonaws.com/reference.png",
        role: "reference_image"
      }
    ]
  });
const requestId = submission.id;

// Poll
let result;
while (true) {
  result = await client.videos.generations.retrieve(requestId);

  if (result.status === "completed") break;
  if (result.status === "failed") {
    throw new Error(result.error.message);
  }

  await new Promise((r) => setTimeout(r, 5000));
}

console.log(result.data[0].url);

MiniMax H3 pricing

Pay per video. No subscriptions, no minimums, no idle GPU costs.

MiniMax H3

$0.15 /second

/v2/videos/generations

MiniMax H3 video generation at 2K resolution with keyframe and reference image control.

ResolutionPrice5s videos for $10
2K$0.15 /second13

MiniMax H3 API — frequently asked questions

How do I use the MiniMax H3 API?

Install the official Node.js (npm install getimg-ai) or Python (pip install getimg) SDK, or call the endpoint directly over HTTP from any client.

Make a POST request to /v2/videos/generations with your API key and a JSON body containing the model ID (minimax-h3) and a prompt. The API returns a request ID. Poll the status endpoint until the video is ready.

See the API documentation for the full parameter reference.

Which SDK should I use?

Use getimg-ai for Node.js and TypeScript projects, and getimg for Python. Both wrap the same REST API with typed inputs, retries, and built-in polling for video generations.

For other languages — Go, Rust, Java, Ruby, PHP — call the HTTP endpoints directly. Every example on this page includes a cURL snippet you can port to any HTTP client.

What resolutions does MiniMax H3 support?

MiniMax H3 supports the following resolutions: 2K across 6 aspect ratios (21:9, 16:9, 4:3, 1:1, 3:4, 9:16). Videos can be 5 to 15 seconds long.

How much does MiniMax H3 cost?

MiniMax H3 uses pay-per-video pricing with no subscriptions or minimum commitments. Rates vary by resolution and sound options. See the pricing table above for exact rates per second.

Does MiniMax H3 support reference images?

Yes. MiniMax H3 accepts up to 1 last frame and up to 1 first frame and up to 9 reference images. Reference images guide the generation toward a specific style, composition, or subject. See the example above.

Start building with MiniMax H3

Grab an API key, install the SDK, ship video generation today.

Get your API keyRead the docs

Looking for other models? View all available models