> ## Documentation Index
> Fetch the complete documentation index at: https://comfyuiwiki.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Use MiniMax H3 with Comfy Router

> Call minimax/minimax-h3 through Comfy Router: endpoint, request shape, and the MP4 with native stereo audio that Router returns

API Reference for MiniMax H3. MiniMax H3 (Hailuo 03) is an omni-modal video model. It generates the picture and its audio track together in one pass, so a finished clip already carries dialogue, sound effects and music.

## Quick start

Create a key in [your Comfy workspace](https://platform.comfy.org/profile/api-keys?onboarding=router) and export it as `COMFY_API_KEY`. The Python and TypeScript snippets use the Comfy SDKs (`pip install comfy-sdk` and `npm install @comfyorg/sdk`); the cURL snippet is the same call over raw HTTP.

**Model ID:** `minimax/minimax-h3`

**Endpoint:** `POST https://api.comfy.org/v2/models/minimax/minimax-h3`

<Tabs defaultTabIndex={1}>
  <Tab title="Wait for the result">
    <CodeGroup>
      ```python Python theme={null}
      import asyncio
      from comfy_sdk import AsyncComfy

      # Reads COMFY_API_KEY from the environment.
      # The SDK automatically creates an idempotency key and reuses it for automatic retries.
      async def main():
          async with AsyncComfy() as client:
              result = await client.models.run(
                  "minimax/minimax-h3",
                  {
                      "content": [
                          {
                              "text": "A single red maple leaf resting on a plain white background.",
                              "type": "text",
                          },
                      ],
                      "duration": 5,
                      "ratio": "16:9",
                      "resolution": "768P",
                  },
              )

          print("video:", result["task"]["content"]["url"])

      asyncio.run(main())
      ```

      ```typescript TypeScript theme={null}
      import { comfy } from "@comfyorg/sdk";

      // Reads COMFY_API_KEY from the environment.
      // The SDK automatically creates an idempotency key and reuses it for automatic retries.
      type Result = { task: { content: { url: string } } };
      const result = await comfy.models.run<Result>("minimax/minimax-h3", {
        content: [
          {
            text: "A single red maple leaf resting on a plain white background.",
            type: "text",
          },
        ],
        duration: 5,
        ratio: "16:9",
        resolution: "768P",
      });
      if (result.kind !== "json") throw new Error("expected a JSON result");

      console.log("video:", result.data.task.content.url);
      ```

      ```bash cURL theme={null}
      curl https://api.comfy.org/v2/models/minimax/minimax-h3 \
        -H "X-API-Key: $COMFY_API_KEY" \
        -H "Idempotency-Key: $(uuidgen)" \
        -H "Content-Type: application/json" \
        -d "{\"content\": [{\"text\":\"A single red maple leaf resting on a plain white background.\",\"type\":\"text\"}], \"duration\": 5, \"ratio\": \"16:9\", \"resolution\": \"768P\"}"
      ```
    </CodeGroup>
  </Tab>

  <Tab title="Queue and collect later">
    The same body, sent to `POST https://api.comfy.org/v2/models/minimax/minimax-h3/requests`. Router answers `201` with a `request_id` as soon as the run is admitted, and the result is collected once it is ready, from this process or another one. [Queued delivery](/development/comfy-router/queue) walks through status, cancellation and collection.

    <CodeGroup>
      ```python Python theme={null}
      import asyncio
      from comfy_sdk import AsyncComfy

      # Reads COMFY_API_KEY from the environment.
      # Each submit() call mints its own Idempotency-Key and reuses it for automatic retries.
      async def main():
          async with AsyncComfy() as client:
              handle = await client.models.submit(
                  "minimax/minimax-h3",
                  {
                      "content": [
                          {
                              "text": "A single red maple leaf resting on a plain white background.",
                              "type": "text",
                          },
                      ],
                      "duration": 5,
                      "ratio": "16:9",
                      "resolution": "768P",
                  },
              )
              print("request_id:", handle.request_id)  # with the model ID, all another process needs

              # Poll until the request completes, waiting the Retry-After the server names.
              async for update in handle.iter_events():
                  print(update.status, update.queue_position)

              # The provider's own payload, the same value models.run() returns.
              # A request that failed or was cancelled raises the typed Router error here.
              result = await handle.get()

          print("video:", result["task"]["content"]["url"])

      asyncio.run(main())
      ```

      ```typescript TypeScript theme={null}
      import { comfy } from "@comfyorg/sdk";

      // Reads COMFY_API_KEY from the environment.
      // Each submit() call mints its own Idempotency-Key and reuses it for automatic retries.
      type Result = { task: { content: { url: string } } };
      const handle = await comfy.models.submit<Result>("minimax/minimax-h3", {
        content: [
          {
            text: "A single red maple leaf resting on a plain white background.",
            type: "text",
          },
        ],
        duration: 5,
        ratio: "16:9",
        resolution: "768P",
      });
      console.log("requestId:", handle.requestId); // with the model ID, all another process needs

      // Poll until the request completes, waiting the Retry-After the server names.
      for await (const update of handle.events()) {
        console.log(update.status, update.queuePosition);
      }

      // The same result models.run() returns. A request that failed or was cancelled rejects here.
      const result = await handle.get();
      if (result.kind !== "json") throw new Error("expected a JSON result");

      console.log("video:", result.data.task.content.url);
      ```

      ```bash cURL theme={null}
      # 1. Submit. Router answers 201 with request_id, status_url, response_url and cancel_url.
      curl https://api.comfy.org/v2/models/minimax/minimax-h3/requests \
        -H "X-API-Key: $COMFY_API_KEY" \
        -H "Idempotency-Key: $(uuidgen)" \
        -H "Content-Type: application/json" \
        -d "{\"content\": [{\"text\":\"A single red maple leaf resting on a plain white background.\",\"type\":\"text\"}], \"duration\": 5, \"ratio\": \"16:9\", \"resolution\": \"768P\"}"

      # 2. Poll until status is COMPLETED, waiting the Retry-After seconds each response names.
      REQUEST_ID="<request_id from the 201 body>"
      curl -i https://api.comfy.org/v2/models/minimax/minimax-h3/requests/$REQUEST_ID/status \
        -H "X-API-Key: $COMFY_API_KEY"

      # 3. Collect. 200 with the model's native output, 202 with the status body while it is still running.
      curl https://api.comfy.org/v2/models/minimax/minimax-h3/requests/$REQUEST_ID \
        -H "X-API-Key: $COMFY_API_KEY"
      ```
    </CodeGroup>
  </Tab>
</Tabs>

## Schema

### Input

<ParamField body="aigc_watermark" type="boolean">
  Whether to add an AIGC watermark to the output. Defaults to false.
</ParamField>

<ParamField body="callback_url" type="string">
  Optional. URL to receive task state changes after challenge verification.
</ParamField>

<ParamField body="content" type="object[]" required>
  Content items driving the generation. Must contain one non-empty text item; optionally add first\_frame/last\_frame images or reference\_\* media.
</ParamField>

<ParamField body="content[].audio_url" type="object">
  Audio source. Required for audio\_url items.
</ParamField>

<ParamField body="content[].audio_url.url" type="string">
  A publicly reachable URL, an mm\_file://\{file\_id} reference, or a data URI.
</ParamField>

<ParamField body="content[].image_url" type="object">
  Image source. Required for image\_url items.
</ParamField>

<ParamField body="content[].image_url.url" type="string">
  A publicly reachable URL, an mm\_file://\{file\_id} reference, or a data URI.
</ParamField>

<ParamField body="content[].role" type="string">
  Role of a media item. Options: first\_frame, last\_frame, reference\_image, reference\_video, reference\_audio, base\_video. Keyframe roles and reference\_\* roles are mutually exclusive within a request; base\_video marks the source video of a video regeneration request.
</ParamField>

<ParamField body="content[].text" type="string">
  The prompt text. Exactly one non-empty text item is required per request.
</ParamField>

<ParamField body="content[].type" type="string" required>
  Content item type. Options: text, image\_url, video\_url, audio\_url.
</ParamField>

<ParamField body="content[].video_url" type="object">
  Video source. Required for video\_url items.
</ParamField>

<ParamField body="content[].video_url.url" type="string">
  A publicly reachable URL, an mm\_file://\{file\_id} reference, or a data URI.
</ParamField>

<ParamField body="duration" type="integer" required>
  Video length in seconds, 5 to 15.
</ParamField>

<ParamField body="model" type="string">
  ID of model. Options: MiniMax-H3. Router callers may omit this field or send null; Router injects the model selected by the request path before provider dispatch.
</ParamField>

<ParamField body="ratio" type="string">
  Aspect ratio. Options: adaptive (default), 21:9, 16:9, 4:3, 1:1, 3:4, 9:16. Required and must not be adaptive for text-to-video; ignored (treated as adaptive) for first-frame or last-frame generation.
</ParamField>

<ParamField body="resolution" type="string" required>
  Video resolution. Options: 2K, 768P.
</ParamField>

<ParamField body="seed" type="integer">
  Random seed in \[-1, 2^32 - 1]; omitted or -1 is random.

  Format: `int64`
</ParamField>

Generated from the schema Router serves at `GET /v2/models/minimax/minimax-h3/openapi.json`, the same document it validates a call against before the request reaches the provider.

### Output

<ResponseField name="task" type="object">
  A Minimax V2 video generation task.
</ResponseField>

<ResponseField name="task.content" type="object">
  Generated output; present when status is succeeded.
</ResponseField>

<ResponseField name="task.content.prompt" type="string">
  The enhanced video prompt produced by a succeeded h3\_context\_ir task.
</ResponseField>

<ResponseField name="task.content.url" type="string">
  Time-limited URL of the generated MP4. Query again for a refreshed URL.
</ResponseField>

<ResponseField name="task.duration" type="number">
  The duration of the generated video in seconds.
</ResponseField>

<ResponseField name="task.error" type="object">
  Error details when status is failed; carries code and message.
</ResponseField>

<ResponseField name="task.id" type="string">
  The task ID.
</ResponseField>

<ResponseField name="task.model" type="string">
  The model used for the task.
</ResponseField>

<ResponseField name="task.ratio" type="string">
  The actual aspect ratio of the generated video.
</ResponseField>

<ResponseField name="task.resolution" type="string">
  The resolution of the generated video.
</ResponseField>

<ResponseField name="task.status" type="string">
  Task status. Options: queued, running, succeeded, failed, cancelled, expired.
</ResponseField>

<ResponseField name="task.task_type" type="string">
  The type of the task.
</ResponseField>

<ResponseField name="task.usage" type="object">
  Usage recorded for the task.
</ResponseField>

<ResponseField name="task.usage.completion_tokens" type="integer" />

<ResponseField name="task.usage.input_image_count" type="integer" />

<ResponseField name="task.usage.input_seconds" type="number" />

<ResponseField name="task.usage.output_seconds" type="number" />

<ResponseField name="task.usage.prompt_tokens" type="integer" />

<ResponseField name="task.usage.total_seconds" type="number" />

<ResponseField name="task.usage.total_tokens" type="integer" />

## Examples

### Input

```json theme={null}
{
  "content": [
    {
      "text": "A single red maple leaf resting on a plain white background.",
      "type": "text"
    }
  ],
  "duration": 5,
  "ratio": "16:9",
  "resolution": "768P"
}
```

### Output

```json theme={null}
{
  "task": {
    "content": {
      "url": "https://example.invalid/minimax/minimax-h3/generated.mp4"
    },
    "duration": 6,
    "id": "3f7a1b28-5c0d-4e91-8a6f-1b2c3d4e5f60",
    "model": "MiniMax-H3",
    "ratio": "16:9",
    "resolution": "768P",
    "status": "succeeded",
    "usage": {
      "output_seconds": 6,
      "total_seconds": 6
    }
  }
}
```

**The MP4 carries a generated audio track.** H3 is omni-modal: voice, sound effects and music are synthesised jointly with the picture in a single forward pass, not layered on afterwards, and the result is one MP4 with native stereo audio. Nothing in the request body above switches that off: the documented fields are the prompt content, `duration`, `ratio`, `resolution`, `seed`, `aigc_watermark` and `callback_url`, and none of them selects a silent render. A pipeline that means to lay its own voiceover over the clip therefore has to mute or strip the returned track first, or it will mux over a clip that is already talking. The prompt is what steers the audio: describe the dialogue, the sound effects and the music in the same prompt block as the shot. See the [MiniMax H3 prompt guide](/tutorials/video/minimax/minimax-h3-prompt-guide) for how to structure one, and the [MiniMax H3 overview](/tutorials/video/minimax/minimax-h3) for what the model does in ComfyUI.

The video URL expires. Router re-hosts the MP4 and answers a Comfy-signed URL valid for 12 hours, falling back to MiniMax's own shorter-lived link when the re-host fails. The `task.content.url` description in the response schema above tells you to query again for a refreshed URL: that is MiniMax's own wording for their API, carried through unchanged, and Router does not expose MiniMax's task-query route. Re-reading a queued request's result route returns the stored result document while the request is [retained, for 24 hours after it completes](/development/comfy-router/queue#idempotency-and-billing), but nothing documents that read as minting a newly signed URL. Download the MP4 promptly rather than storing the link.

## Before you ship

The SDKs create an `Idempotency-Key` and reuse it for automatic retries. For manual retries, reuse the original key. Router can hold the connection for up to 10 minutes.

When a request fails, Router sends an `X-Comfy-Error-Type` response header explaining why. A `422` means Router rejected the input before calling the provider, and a `413` means the request body was larger than Router accepts. Download generated assets promptly because [result URLs can expire](/development/comfy-router/reference#result-assets).

Any size limit named in a field description above is the provider's own bound on that field, quoted from the provider's specification. Router applies a separate cap to the whole request body, which base64-encoded media counts against: see [request body size](/development/comfy-router/limitations#request-bodies-are-capped).

This page documents one partner model called through Comfy Router. The same `comfy-sdk` / `@comfyorg/sdk` package also ships a second client, for running a whole ComfyUI workflow graph on Comfy Cloud: `Comfy(api_key=...)` / `new Comfy({ apiKey })`, with `client.workflows`, `client.assets` and `client.jobs`. See [Comfy SDKs](/development/api-development/sdks).

<CardGroup cols={3}>
  <Card title="Headers" icon="list" href="/development/comfy-router/headers">
    Authentication, idempotency, request IDs, error buckets, retry pacing, spend limits.
  </Card>

  <Card title="Using the Router API" icon="code" href="/development/comfy-router/api">
    Model discovery, validation errors, retries, and billing.
  </Card>

  <Card title="Limitations" icon="triangle-exclamation" href="/development/comfy-router/limitations">
    What Router does not do today, and what to use instead.
  </Card>
</CardGroup>