Kling Video

Create text-to-video, image-to-video, and Omni reference/edit video tasks with Kling.

Endpoints

POST /text-to-video/{model}
POST /image-to-video/{model}
POST /omni-video/{model}
Authorization: Bearer $API_KEY
Content-Type: application/json

{model} is the complete public Kling video model ID, for example kling/kling-4.0. The upstream accepts the task immediately; save the returned task ID and poll via GET /v1/tasks/{taskId}.

Models and supported endpoints

Model	text-to-video	image-to-video	omni-video	Resolutions	Duration
`kling/kling-4.0`	✓	✓	✓	720p, 1080p	3–30s
`kling/kling-4.0-flash`	✓	✓	✓	720p, 1080p	3–30s
`kling/kling-3.0-omni`	—	—	✓	720p, 1080p, 4k	3–15s
`kling/kling-3.0-turbo`	✓	✓	—	720p, 1080p	3–15s
`kling/kling-3.0`	✓	✓	—	720p, 1080p, 4k	3–15s

Text to video

Applies to: kling-4.0, kling-4.0-flash, kling-3.0-turbo, kling-3.0

Field	Type	Required	Description
`prompt`	string	Yes	Generation prompt, up to 3072 characters.
`settings.resolution`	string	No	Default `1080p` (4.0 series) or `720p` (3.0 series).
`settings.aspect_ratio`	string	No	`21:9`, `16:9` (default), `9:16`, `1:1`. The 3.0 series does not support `21:9`.
`settings.duration`	integer	No	Seconds; reads explicit duration from prompt when omitted, otherwise defaults to 10.
`settings.audio`	string	No	`native` (generate audio) or `off`; `kling-3.0-turbo` defaults to `native`, others default to `off`.
`options.callback_url`	string	No	Public terminal callback URL.
`options.external_task_id`	string	No	Custom task ID, unique within the account; does not overwrite the system ID.

curl "https://api.modelrover.ai/text-to-video/kling/kling-4.0" \
-H "Authorization: Bearer $API_KEY" \
-H "Content-Type: application/json" \
-d '{
  "prompt": "A girl sitting on a train, gazing out the window with a wistful expression, her head swaying gently with the carriage",
  "settings": {
    "resolution": "1080p",
    "aspect_ratio": "16:9",
    "duration": 10,
    "audio": "native"
  },
  "options": {
    "callback_url": "https://example.com/hooks/video?token=unguessable"
  }
}'

Image to video

Applies to: kling-4.0, kling-4.0-flash, kling-3.0-turbo, kling-3.0

Provide reference frames through the contents array. Each item requires a type and the corresponding text, url, or id field.

`type`	Description	Applies to
`prompt`	Prompt text (exactly one required)	All
`first_frame`	First-frame image (URL or base64); at least one frame is required	All
`last_frame`	Last-frame image	All
`middle_frame`	Mid-video keyframe; multiple allowed, ordered sequentially	4.0 series only

Images must be .jpg / .jpeg / .png, ≤ 30 MB each, ≥ 300px on each side, aspect ratio between 2:5 and 5:2, up to 10 total. When multiple middle_frame items are provided, assign each an id and reference them with @id in the prompt.

{
  "contents": [
    { "type": "first_frame", "url": "https://example.com/start.jpg", "id": "frame_1" },
    { "type": "last_frame",  "url": "https://example.com/end.jpg",   "id": "frame_2" },
    { "type": "prompt", "text": "Smooth cinematic transition from the dawn forest in @frame_1 to the starlit snowy forest in @frame_2" }
  ],
  "settings": { "resolution": "1080p", "duration": 8 },
  "options": { "callback_url": "https://example.com/hooks/video?token=unguessable" }
}

Omni video

Applies to: kling-4.0, kling-4.0-flash, kling-3.0-omni

The omni endpoint extends image-to-video with reference images and videos, elements, and voices, enabling reference-to-video, video editing, and other composite scenarios.

`type`	Description
`prompt`	Prompt text (exactly one required)
`first_frame` / `middle_frame` / `last_frame`	Keyframe images; same rules as image-to-video
`refer_image`	Style or content reference image, up to 10; cannot be mixed with keyframes
`feature_video`	Feature reference video (motion or style reference), up to 5
`base_video`	Video to edit, at most 1
`element`	Preset or custom element, requires `element_id`; up to 7 total (≤ 3 video elements)
`voice`	Custom voice, requires `voice_id`; up to 3; 4.0 series only

Videos must be .mp4 or .mov, ≤ 300 MB each, 3–30s each; images, videos, elements, and voices combined cannot exceed 15.

{
  "contents": [
    { "type": "refer_image", "url": "https://example.com/character.jpg", "id": "char_1" },
    { "type": "prompt", "text": "Preserve the facial features and clothing of @char_1 and have her running across a meadow" }
  ],
  "settings": {
    "resolution": "1080p",
    "aspect_ratio": "16:9",
    "duration": 8,
    "audio": "native"
  },
  "options": { "callback_url": "https://example.com/hooks/video?token=unguessable" }
}

settings.audio additionally accepts original on the omni endpoint, which preserves the source video's audio track. original cannot be combined with a voice item.

Create response

{
  "code": 0,
  "data": {
    "id": "9dcf8ab2-c4b7-42be-a516-597223897b99",
    "status": "submitted"
  }
}

data.id is the platform task ID. Task statuses are submitted, processing, succeeded, and failed.

Task response

curl "https://api.modelrover.ai/v1/tasks/9dcf8ab2-c4b7-42be-a516-597223897b99" \
-H "Authorization: Bearer $API_KEY"

{
  "code": 0,
  "data": {
    "id": "9dcf8ab2-c4b7-42be-a516-597223897b99",
    "status": "succeeded",
    "videos": [
      { "url": "https://cdn.klingai.com/...", "duration": 10 }
    ],
    "usage": { "output_duration": 10 }
  }
}

Video results are temporarily retained for 48 hours and automatically deleted upon expiration—download and transfer them to your own storage immediately after completion.

Restrictions

options.watermark_info, client routing, and paid service-tier selections are rejected.
external_task_id must be unique within the account and cannot be reused.
Body size defaults to 50 MiB; base64-encoded images count toward the body size.

See Tasks and callbacks for polling, callback delivery, and task lifecycle.

Kling Video

On this page