Multimodal Video Generation

Model name. Currently available value: MiniMax-H3.

Multimodal input content array describing the information used to generate the video. Each item is distinguished by type (text / image_url / video_url / audio_url) and can be labeled with role to indicate its purpose.

Every request must include one non-empty text item (prompt required); omitting it returns a parameter error.

Supported input combinations (for different generation scenarios):

• Text-to-video: one text item only.
• Image-to-video—first frame: text + 1 image_url (role=first_frame or omitted).
• Image-to-video—last frame: text + 1 image_url (role=last_frame).
• Image-to-video—first and last frames: text + 2 image_url items (role values first_frame and last_frame, respectively).
• Multimodal reference-to-video: text + a combination of reference images (role=reference_image), reference videos (role=reference_video), and reference audio (role=reference_audio).

Image-to-video and multimodal reference-to-video are mutually exclusive: if content includes any of the reference_image / reference_video / reference_audio roles, it cannot include first_frame / last_frame (and vice versa). The two cannot be mixed.

Input media limits (total request body size ≤ 64 MB; use public URLs for large files instead of Base64)

Images (image_url):

Videos (video_url) (multimodal reference scenarios only):

Audio (audio_url) (multimodal reference scenarios only):

The format is as follows:

[
  {
    "type": "text",
    "text": "string"
  },
  {
    "type": "image_url",
    "image_url": {
      "url": "string"
    },
    "role": "first_frame"
  },
  {
    "type": "video_url",
    "video_url": {
      "url": "string"
    },
    "role": "reference_video"
  },
  {
    "type": "audio_url",
    "audio_url": {
      "url": "string"
    },
    "role": "reference_audio"
  }
]

{
  "type": "text",
  "text": "string"
}

{
  "type": "image_url",
  "image_url": {
    "url": "string"
  },
  "role": "first_frame"
}

{
  "type": "video_url",
  "video_url": {
    "url": "string"
  },
  "role": "reference_video"
}

{
  "type": "audio_url",
  "audio_url": {
    "url": "string"
  },
  "role": "reference_audio"
}

Video resolution. Currently available values: 768P, 2K.

Generated video duration (seconds), required and must be an integer.

Available values: 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15

Aspect ratio of the generated video. Defaults to adaptive (automatically selects the most appropriate aspect ratio based on the input; the actual ratio is available from the ratio field in the query API).

Available values: adaptive, 21:9, 16:9, 4:3, 1:1, 3:4, 9:16.

Text-to-video (t2va, content contains only text): ratio is required and cannot be adaptive; available values are 21:9, 16:9, 4:3, 1:1, 3:4, 9:16.

Image-to-video (i2va, content contains first_frame / last_frame images): the aspect ratio is determined by the input image, and ratio is always adaptive; passing another valid value does not return an error, but it is ignored and processed as adaptive.

Multimodal reference-to-video (r2va, content contains reference_image / reference_video / reference_audio): ratio is optional and defaults to adaptive; any specific ratio listed above can also be explicitly set.

Callback URL for task status change notifications. Once configured, the server first sends a verification request containing a challenge field (return the challenge unchanged within 3 seconds to complete verification). After verification succeeds, it sends a POST push notification to this URL whenever task status changes. The push body has the same structure as the response from the Query Task API.

Callback status values: queued (queued), running (running), succeeded (succeeded), failed (failed), cancelled (cancelled).

Whether to add an AIGC watermark to the generated video. Defaults to false.

Task status:

• queued: Queued
• running: Running
• succeeded: Succeeded
• failed: Failed
• cancelled: Cancelled

The recommended polling interval is 3–5 seconds, until status is succeeded, failed, or cancelled.

Error information. Not returned when the task succeeds; returned when the task fails.

Task output content, returned after the task succeeds.

Billable usage for this request. Video tasks return fields measured in seconds.