# H3 Max Reference to Video API

> H3 Max video generation that keeps subjects consistent using reference images, videos and audio clips

Generates video guided by reference assets: up to 9 images, 3 videos and 3 audio clips, with at most 12 files in total. The prompt refers to them as Image 1, Video 1, Audio 1 and so on. Optional first, middle (`middle_image_url` at `middle_frame_time`) and last frames pin the start, middle and end. `duration` is 0.92 to 15 seconds, and `aspect_ratio` can be adaptive.

- Model id: `minimax/h3-max/reference-to-video`
- Made by: MiniMax
- Category: Reference to Video
- Output: video
- Price per second of your output video: from $0.05 / second of output video (cheapest run $0.05 (50 API credits))
- Run: `POST https://api.doitong.com/v1/run/minimax/h3-max/reference-to-video`
- Quote: `POST https://api.doitong.com/v1/quote/minimax/h3-max/reference-to-video`
- Docs page: https://doitong.com/developers/models/minimax/h3-max/reference-to-video

## Pricing

Price: **from $0.05 / second of output video** (the lowest tier; the quote prices your exact input). Price per second of your output video: the input files are measured before the run (length, resolution, frame rate). The quote measures the same files and is exactly what the run charges.

Price by tier:

- 480P: $0.05 / second of output video
- 768P: $0.08 / second of output video
- 1080P: $0.16 / second of output video

The full example input below costs $0.40 = 400 API credits (5 s output video × $0.08 (768P)).

Prices are the upstream price in USD, paid from the API balance in API credits (1000 API credits = 1 USD). `POST https://api.doitong.com/v1/quote/minimax/h3-max/reference-to-video` with the same JSON body returns the exact price for that input; the quote is exactly what a run charges. Invalid input is never charged and failed or canceled requests are refunded automatically.

## Input parameters

Send either the canonical name or the native field name (not both). Canonical names are the same across models.

| Parameter | Canonical name | Type | Required | Default | Allowed values / range | Description |
|---|---|---|---|---|---|---|
| `prompt` |  | string | yes |  | 1 – 50000 characters | Video-generation instructions with references identified by asset type and position within each reference list, using labels such as Audio 1, Video 1, Image 1, Image 2, and so on. |
| `reference_image_urls` | `images` | array of image URL | no |  | ≤ 9 items | Subject or style reference image URLs, sharing a 12-file maximum with all video and audio references. Identify these images in the prompt by list position using Image 1, Image 2, and subsequent labels. |
| `reference_video_urls` |  | array of video URL | no |  | ≤ 3 items | Motion or other reference video URLs, identified by list position in the prompt using Video 1, Video 2, and subsequent labels. The shared allowance across image, video, and audio references is 12 files; video clips must each last 2–15 seconds and total no more than 15 seconds. |
| `reference_audio_urls` |  | array of audio URL | no |  | ≤ 3 items | Supply reference audio as URLs and identify the clips in the prompt using Audio 1, Audio 2, and subsequent labels. Each clip must be 2–15 seconds long, with no more than 15 seconds of audio in total. Any of the image, video, and audio reference types can be supplied alone or combined, subject to a shared maximum of 12 files. |
| `image_url` | `image` | string (image URL) \| null | no |  | ≥ 1 characters | Optional image URL defining the exact opening frame and determining the output canvas. Reference assets maintain subject consistency. |
| `middle_image_url` |  | string (image URL) \| null | no |  | ≥ 1 characters | Optional guiding image for middle_frame_time, available only with native 480P or 768P generation and with both a starting image and an ending image supplied. |
| `middle_frame_time` |  | number \| null | no |  | > 0, < 15 | Middle-image placement time, usable only with middle_image_url and restricted to positions strictly inside the requested video's first and last frames. Supply seconds elapsed since the start; the value is rounded to the nearest 24 fps frame. |
| `end_image_url` | `end_image` | string (image URL) \| null | no |  | ≥ 1 characters | Optional final-image URL that fixes the video's exact ending frame to the supplied image. |
| `duration` |  | number | no | `5` | 0.92 – 15 | Requested duration in seconds, allowing an actual video length up to approximately 0.7 s beyond that request. |
| `resolution` |  | enum (string) | no | `768P` | `480P`, `768P`, `1080P` | Generation resolution selection: 1080P requires latent refinement of material generated natively at 768P, rather than native generation at the selected resolution. |
| `seed` |  | integer \| null | no |  |  | Seed for random generation; if omitted, one is chosen randomly. |
| `enable_safety_checker` |  | boolean | no | `true` |  | Safety checker activation; true enables checking. |
| `prompt_expansion_mode` |  | string | yes | `balanced` |  | Pre-generation prompt rewriting effort: 'quality' develops a richer prompt over as much as approximately 30s, while 'balanced' returns in roughly a second. Choose 'disabled' to omit prompt expansion. |
| `aspect_ratio` |  | enum (string) | no | `adaptive` | `adaptive`, `21:9`, `16:9`, `4:3`, `1:1`, `3:4`, `9:16` | Aspect ratio for the output video. |
| `reference_conditions` |  | array of object {type, url} | no |  | ≤ 12 items | Reference list supplied as an alternative to reference_*_urls, never alongside them. Ordering across modalities carries meaning, so the sequence is maintained throughout processing. |
| `reference_conditions[].type` |  | enum (string) | yes |  | `image`, `video`, `audio` | Modality of the reference; ordering across modalities in the list carries meaning. |
| `reference_conditions[].url` |  | string | yes |  | ≥ 1 characters | Reference media's URL. |

File parameters (`images`, `reference_video_urls`, `reference_audio_urls`, `image`, `middle_image_url`, `end_image`) take a public `https://` URL, a `data:` URI up to 10 MB, or the `file_url` of an upload from `POST /files`.

## Example input

Required parameters only:

```json
{
  "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field",
  "prompt_expansion_mode": "balanced"
}
```

A fuller example:

```json
{
  "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field",
  "prompt_expansion_mode": "balanced"
}
```

## Run it

Set `DOITONG_API_KEY` to a key from https://doitong.com/developers/dashboard#keys. `?wait=20` returns the finished request in the same call when it is ready within 20 seconds; otherwise poll `GET /requests/{id}` (it also accepts `?wait=`), or pass `webhook_url`.

### curl

```bash
# Run the model; wait up to 20 s for the finished result in the same call
curl -X POST "https://api.doitong.com/v1/run/minimax/h3-max/reference-to-video?wait=20" \
  -H "Authorization: Bearer $DOITONG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field",
    "prompt_expansion_mode": "balanced"
  }'

# Still "queued" or "running"? Poll with the id from the response:
curl "https://api.doitong.com/v1/requests/REQUEST_ID?wait=20" \
  -H "Authorization: Bearer $DOITONG_API_KEY"
```

### JavaScript

```javascript
const API = "https://api.doitong.com/v1";
const headers = {
  "Authorization": "Bearer " + process.env.DOITONG_API_KEY,
  "Content-Type": "application/json",
};

// Run the model; wait up to 20 s for the finished result in the same call
let res = await fetch(API + "/run/minimax/h3-max/reference-to-video?wait=20", {
  method: "POST",
  headers,
  body: JSON.stringify({
    "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field",
    "prompt_expansion_mode": "balanced"
  }),
});
let request = await res.json();
if (!res.ok) throw new Error(request.error.code + ": " + request.error.message);

// Not finished yet: long-poll until it is
while (request.status === "queued" || request.status === "running") {
  res = await fetch(API + "/requests/" + request.id + "?wait=20", { headers });
  request = await res.json();
}
if (request.status !== "succeeded") throw new Error(JSON.stringify(request.error));
console.log(request.outputs.map((o) => o.url), request.cost.credits + " API credits ($" + request.cost.usd + ")");
```

### Python

```python
import os
import requests

API = "https://api.doitong.com/v1"
headers = {"Authorization": "Bearer " + os.environ["DOITONG_API_KEY"]}

payload = {
    "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field",
    "prompt_expansion_mode": "balanced",
}

# Run the model; wait up to 20 s for the finished result in the same call
r = requests.post(API + "/run/minimax/h3-max/reference-to-video", params={"wait": 20}, json=payload, headers=headers)
request = r.json()
if not r.ok:
    raise RuntimeError(request["error"]["code"] + ": " + request["error"]["message"])

# Not finished yet: long-poll until it is
while request["status"] in ("queued", "running"):
    request = requests.get(API + "/requests/" + request["id"], params={"wait": 20}, headers=headers).json()
if request["status"] != "succeeded":
    raise RuntimeError(request["error"])
print([o["url"] for o in request["outputs"]], request["cost"]["credits"], "API credits", "$" + request["cost"]["usd"])
```

### MCP

The MCP server at `https://api.doitong.com/v1/mcp` exposes the tools `search_models`, `get_model`, `quote`, `run`, `get_request`, `cancel_request`, `list_requests`, `create_upload` and `get_account`. A `run` call for this model:

```json
{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "run",
    "arguments": {
      "model_id": "minimax/h3-max/reference-to-video",
      "input": {
        "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field",
        "prompt_expansion_mode": "balanced"
      },
      "wait_seconds": 20
    }
  }
}
```

## Response

Every call returns a request object. `status` is one of `queued`, `running`, `succeeded`, `failed`, `canceled`. `outputs` lists the generated files with URLs on our storage (kept forever, or 7 days with `?storage=temporary`; `storage` says which); `output` is the model's own result object with the same URLs.

```json
{
  "id": "req_8k2m4q7x1z9c3v5b6n0p2r4t",
  "object": "request",
  "model": "minimax/h3-max/reference-to-video",
  "status": "succeeded",
  "created_at": "2026-10-20T12:00:00Z",
  "started_at": "2026-10-20T12:00:02Z",
  "completed_at": "2026-10-20T12:00:09Z",
  "queue_position": null,
  "cost": {
    "credits": 400,
    "usd": "0.40",
    "charged": 400,
    "refunded": 0
  },
  "input": {
    "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field",
    "prompt_expansion_mode": "balanced"
  },
  "output": {
    "video": {
      "url": "https://content.doitong.com/dev-api/2026/10/20/req_8k2m4q7x1z9c3v5b6n0p2r4t/0-output.mp4",
      "content_type": "video/mp4",
      "file_name": "0-output.mp4",
      "file_size": 1843200
    },
    "expanded_prompt": "string",
    "seed": 42,
    "timings": {}
  },
  "outputs": [
    {
      "type": "video",
      "url": "https://content.doitong.com/dev-api/2026/10/20/req_8k2m4q7x1z9c3v5b6n0p2r4t/0-output.mp4",
      "content_type": "video/mp4",
      "file_size": 1843200,
      "width": 1280,
      "height": 720,
      "duration": 5,
      "stored": true
    }
  ],
  "error": null,
  "metrics": {
    "inference_seconds": 6.8
  },
  "urls": {
    "self": "https://api.doitong.com/v1/requests/req_8k2m4q7x1z9c3v5b6n0p2r4t",
    "cancel": "https://api.doitong.com/v1/requests/req_8k2m4q7x1z9c3v5b6n0p2r4t/cancel"
  }
}
```

### Output fields (`output`)

| Field | Type | Description |
|---|---|---|
| `video` | object {url, content_type, file_name, file_size} | Video produced by the model. |
| `expanded_prompt` | string \| null | Model-ready expanded prompt, or null if no expanded text is available. The null cases are disabled expansion, expansion that made no change, and expansion handled internally by the model's hosted API. |
| `seed` | integer | Base seed used to reproduce the generated result. |
| `timings` | object \| null | Timing measurements expressed in seconds, available only on routes reporting backend timings; otherwise null. The 'inference' entry measures how long DiT denoising took on the GPU backend. |

## Related models

- [MiniMax H3 Max Text to Video](https://doitong.com/developers/models/minimax/h3-max/text-to-video.md) (`minimax/h3-max/text-to-video`): from $0.0325 / second of output video
- [H3 Max Camera Controls](https://doitong.com/developers/models/minimax/h3-max/camera-controls.md) (`minimax/h3-max/camera-controls`): from $0.0325 / second
- [H3 Max Image to Video](https://doitong.com/developers/models/minimax/h3-max/image-to-video.md) (`minimax/h3-max/image-to-video`): from $0.0325 / second of output video
- [H3 Max](https://doitong.com/developers/models/minimax/h3-max/insert-video.md) (`minimax/h3-max/insert-video`): from $0.05 / second
- [H3 Max Extend Video](https://doitong.com/developers/models/minimax/h3-max/extend-video.md) (`minimax/h3-max/extend-video`): from $0.05 / second of output video
- [H3 Max Recast](https://doitong.com/developers/models/minimax/h3-max/recast.md) (`minimax/h3-max/recast`): from $0.30 / second of input video
- [H3 Max Reframe](https://doitong.com/developers/models/minimax/h3-max/reframe-video.md) (`minimax/h3-max/reframe-video`): from $0.05 / second of input video
- [H3 Max Relight](https://doitong.com/developers/models/minimax/h3-max/relight.md) (`minimax/h3-max/relight`): from $0.05 / second of input video

## Links

- HTML page: https://doitong.com/developers/models/minimax/h3-max/reference-to-video
- All models and guides for AI agents: https://doitong.com/developers/llms.txt
- OpenAPI 3.1 with this model's typed request body: https://doitong.com/developers/openapi.json?model=minimax/h3-max/reference-to-video
- Errors: https://doitong.com/developers/docs/errors
- Model details as JSON: https://api.doitong.com/v1/models/minimax/h3-max/reference-to-video