> ## Documentation Index > Fetch the complete documentation index at: https://docs.mixroute.ai/llms.txt > Use this file to discover all available pages before exploring further. # MiniMax H3 > MiniMax H3 video generation parameters, multimodal inputs, and output settings. `MiniMax-H3` supports text-to-video, frame-based image-to-video, and multimodal reference generation. Keep `model` and `prompt` at the root and place the provider's content and output controls in `metadata`. `POST https://api.mixroute.ai/v1/video/generations` ## Request Parameters | Field | Type | Required | Description | | ----------------------- | --------- | ----------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `model` | string | Yes | Use the case-sensitive MixRoute model ID `MiniMax-H3`. | | `prompt` | string | Yes | MixRoute prompt, consistent with the text in `metadata.content`. | | `metadata` | object | Yes | Container for the required provider fields below. | | `metadata.content` | object\[] | Yes | Multimodal input list. Every request must contain at least one nonempty `text` item. Text-only generation uses one text item; image/video/audio inputs must follow the mode rules below. | | `metadata.resolution` | string | Yes | Output resolution: `768P` or `2K`, with exact capitalization. Supply an explicit value. | | `metadata.duration` | integer | Yes | Output duration in whole seconds, from 4 through 15 inclusive. Supply an explicit value. | | `metadata.ratio` | string | Conditional | Values: `adaptive`, `21:9`, `16:9`, `4:3`, `1:1`, `3:4`, `9:16`. Default `adaptive` except that text-only generation must explicitly choose a concrete ratio. Frame-based generation always uses the input image ratio. | | `metadata.callback_url` | string | No | Callback URL for task-status changes. The first request carries `challenge`, which must be returned unchanged within 3 seconds. Subsequent POST notifications use the provider task-query schema, with statuses `queued`, `running`, `succeeded`, `failed`, or `cancelled`. | ## Content Items | Field | Type | Required | Description | | ---------------------------------- | ------ | ----------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `metadata.content[].type` | string | Yes | `text`, `image_url`, `video_url`, or `audio_url`. | | `metadata.content[].text` | string | Conditional | Required for text items; nonempty, at most 7000 characters per item. | | `metadata.content[].image_url.url` | string | Conditional | Required for image items. Public URL, accessible `mm_file://{file_id}`, or `data:image/;base64,...` with lowercase format. | | `metadata.content[].video_url.url` | string | Conditional | Required for video references. Public URL, accessible `mm_file://{file_id}`, or `data:video/mp4;base64,...`. | | `metadata.content[].audio_url.url` | string | Conditional | Required for audio references. Public URL, accessible `mm_file://{file_id}`, or `data:audio/;base64,...` with lowercase format. | | `metadata.content[].role` | string | Conditional | Use `first_frame` or `last_frame` for frame-based generation; a single image without a role defaults to first frame. Reference inputs use `reference_image`, `reference_video`, or `reference_audio`, matching the content type. | ## Input Modes | Mode | `metadata.content` | `metadata.ratio` | | --------------------- | ---------------------------------------------------- | -------------------------------------------------------------------- | | Text-to-video | `text` | Required concrete ratio; `adaptive` is not allowed. | | First frame | `text` + `image_url` (`first_frame`) | Always `adaptive`; another valid ratio is ignored. | | Last frame | `text` + `image_url` (`last_frame`) | Always `adaptive`, matching the image. | | First and last frames | `text` + 2 `image_url` (`first_frame`, `last_frame`) | Always `adaptive`; another valid ratio is ignored. | | Reference-to-video | `text` + `reference_*` | Optional; defaults to `adaptive`. A concrete ratio may be specified. | Frame roles and reference roles are mutually exclusive. Provider file IDs must be accessible to the upstream account; they are not MixRoute task IDs. This page covers `MiniMax-H3`, not other H3 variants. ## Media Limits | Media | Constraints | | --------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Image | JPG/JPEG/PNG/WebP/HEIC/HEIF; at most 30 MB each; dimensions 256-5760 px; aspect ratio 0.4-2.5. At most one first frame, one last frame, or nine reference images. | | Reference video | MP4/MOV; H.264/H.265 video and AAC/MP3 audio. At most 50 MB each, three clips, 2-15 s per clip and at most 15 s total; dimensions 256-5760 px, ratio 0.4-2.5, and frame rate 23.976-60 fps. | | Reference audio | WAV/MP3; at most 15 MB each, three clips, 2-15 s per clip and at most 15 s total. | | Request | At most 64 MB including Base64 overhead. Prefer public URLs for large inputs. | ## Examples Set `MIXROUTE_API_KEY` before calling the API. Replace media placeholders with accessible inputs. Accepted requests create billable generation tasks; do not automatically resubmit after a submission timeout. ```bash theme={null} curl --location "https://api.mixroute.ai/v1/video/generations" \ --header "Authorization: Bearer $MIXROUTE_API_KEY" \ --header "Content-Type: application/json" \ --data '{ "model": "MiniMax-H3", "prompt": "A blue circle gently pulses on a clean white background.", "metadata": { "content": [ { "type": "text", "text": "A blue circle gently pulses on a clean white background." } ], "resolution": "768P", "duration": 4, "ratio": "9:16" } }' ``` ```bash theme={null} curl --location "https://api.mixroute.ai/v1/video/generations" \ --header "Authorization: Bearer $MIXROUTE_API_KEY" \ --header "Content-Type: application/json" \ --data '{ "model": "MiniMax-H3", "prompt": "A professor is teaching a class in a classroom.", "metadata": { "content": [ { "type": "text", "text": "A professor is teaching a class in a classroom." }, { "type": "image_url", "image_url": { "url": "https://static.freebeatfit.com/aiImageEdit/result/L001510027166320955392/1780060289776-1.png" }, "role": "first_frame" }, { "type": "image_url", "image_url": { "url": "https://static.freebeatfit.com/aiImageEdit/result/M1509975287180300288_0004_1780047826.png" }, "role": "last_frame" } ], "resolution": "768P", "duration": 4, "ratio": "adaptive" } }' ``` ```bash theme={null} curl --location "https://api.mixroute.ai/v1/video/generations" \ --header "Authorization: Bearer $MIXROUTE_API_KEY" \ --header "Content-Type: application/json" \ --data '{ "model": "MiniMax-H3", "prompt": "Preserve the scene in the reference video and follow the reference audio.", "metadata": { "content": [ { "type": "text", "text": "Preserve the scene in the reference video and follow the reference audio." }, { "type": "video_url", "video_url": { "url": "https://example.com/reference.mp4" }, "role": "reference_video" }, { "type": "audio_url", "audio_url": { "url": "https://example.com/reference.mp3" }, "role": "reference_audio" } ], "resolution": "2K", "duration": 4, "ratio": "adaptive" } }' ``` ### Python ```python theme={null} import json import os import requests payload = json.loads(r''' { "model": "MiniMax-H3", "prompt": "A blue circle gently pulses on a clean white background.", "metadata": { "content": [ { "type": "text", "text": "A blue circle gently pulses on a clean white background." } ], "resolution": "768P", "duration": 4, "ratio": "9:16" } } ''') response = requests.post( "https://api.mixroute.ai/v1/video/generations", headers={"Authorization": "Bearer " + os.environ["MIXROUTE_API_KEY"]}, json=payload, timeout=60, ) response.raise_for_status() print(response.json()) ``` ## Task Results Save the MixRoute task ID returned by submission and poll [Query Video Task](/en/api-reference/endpoint/query-video-task). A successful submission creates a task; read the result only after the task reaches a successful terminal state. Response envelopes and result locations vary by route.