Kling 3.0 Omni - Unifically

Generate high-quality AI videos using Kling 3.0 Omni. Supports text-to-video, raw reference images, persistent subject/object elements, start/end frames, video transform, video reference, and multi-shot generation.

Model

kuaishou/kling-3.0-omni-video

Parameters

Parameter	Type	Required	Default	Description
`video_mode`	string	No	-	Input mode. Allowed: `"elements"`, `"start_end_frame"`, `"transform"`, `"video_reference"`. Omit for text-to-video
`prompt`	string	Conditional	-	Text prompt. Mutually exclusive with `multi_shots`
`mode`	string	No	`"pro"`	`"std"` (720p), `"pro"` (1080p), or `"4k"` (4K — $0.30/sec)
`duration`	integer	No	`5`	Video length in seconds (3–15)
`aspect_ratio`	string	No	-	`"1:1"`, `"9:16"`, `"16:9"`, or `"auto"`. Must be `"auto"` when `video_url` is set; rejected otherwise. Omit to let the server choose (`"auto"` if `video_url`, else `"16:9"`)
`native_audio`	boolean	No	`false`	Generate AI audio from the video content
`image_urls`	string[]	No	-	Raw reference image URLs uploaded per request (max 7). No description, no reuse. Use `@Image1`, `@Image2` in `prompt`
`start_image_index`	integer	No	-	1-based index into `image_urls` marking which reference image anchors the first frame. No end-frame equivalent
`elements`	ElementInput30[]	No	-	Persistent subject/object assets. Auto-created on Kling and auto-deleted if submit fails. Use `@Element1`, `@Element2` in `prompt`. Coexists with `image_urls`
`start_frame_url`	string	No	-	First frame image URL (`start_end_frame` mode)
`end_frame_url`	string	No	-	Last frame image URL (`start_end_frame` mode)
`video_url`	string	No	-	Source video URL (`transform` / `video_reference` modes)
`multi_shots`	array	No	-	Multi-shot sequence (2–6 shots, each with `prompt` and `duration`). Mutually exclusive with `prompt`

Modes

Mode	Required inputs	Optional inputs	Notes
omitted (text-to-video)	`prompt` or `multi_shots`	—	Rejects every input field including `elements`, `image_urls`, `start_frame_url`, `video_url`
`elements`	At least one of `image_urls`, `elements`	—	Combined `image_urls` + `elements` cap of 7. `video_url` rejected
`start_end_frame`	`start_frame_url`	`end_frame_url`, `elements`	—
`transform`	`video_url`	`image_urls`, `elements`	Combined `image_urls` + `elements` cap of 4. `aspect_ratio` must be `"auto"`
`video_reference`	`video_url`	`image_urls`, `elements`	Combined `image_urls` + `elements` cap of 4. `aspect_ratio` must be `"auto"`

Text-to-video with elements — set video_mode: "elements" and pass only the elements array (no image_urls required). The omit-video_mode path is strictly for pure text-to-video with zero inputs; any reference (image or element) requires elements mode.

Reference images vs elements

image_urls — raw images uploaded per request. Single-use, no description, no reuse across tasks. Cap 7. Reference in prompt as @Image1, @Image2, …
elements — persistent subject/object assets stored on the Kling account. Auto-created from your input on submit and auto-deleted if the submit fails. Reference in prompt as @Element1, @Element2, …

Both can be used together; they share the per-mode combined cap.

Elements

v3 Omni supports both IMAGE and VIDEO elements. A video_url and any element with type: "video" are mutually exclusive — passing both returns 400 (the video element counts as your video reference).

"elements": [
  {"description": "Woman in red jacket", "type": "image", "image_urls": ["front.jpg", "side.jpg"]},
  {"description": "Man dancing", "type": "video", "video_url": "dance.mp4"}
]

Each element accepts:

Parameter	Type	Default	Description
`description`	string	`""`	Short description of the element (max 100 chars)
`type`	string	`"image"`	Element source type: `"image"` or `"video"`
`image_urls`	string[]	-	Source image URLs for an image element (max 4)
`video_url`	string	-	Source video URL for a video element

Example - Text-to-Video

curl -X POST https://api.unifically.com/v1/tasks \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "kuaishou/kling-3.0-omni-video",
    "input": {
      "prompt": "A majestic eagle soaring through mountain peaks at sunset",
      "duration": 5,
      "mode": "pro"
    }
  }'

Example - With Elements

curl -X POST https://api.unifically.com/v1/tasks \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "kuaishou/kling-3.0-omni-video",
    "input": {
      "video_mode": "elements",
      "prompt": "@Element1 walks through a park at sunset",
      "elements": [
        {"description": "Woman in red jacket", "type": "image", "image_urls": ["https://example.com/front.jpg"]}
      ],
      "duration": 10,
      "mode": "pro"
    }
  }'

Example - Reference Images with Start Frame

curl -X POST https://api.unifically.com/v1/tasks \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "kuaishou/kling-3.0-omni-video",
    "input": {
      "video_mode": "elements",
      "prompt": "@Image2 walks toward @Image1 across a meadow",
      "image_urls": [
        "https://example.com/character.jpg",
        "https://example.com/scene.jpg"
      ],
      "start_image_index": 2,
      "duration": 5,
      "mode": "pro"
    }
  }'

Example - Multi-Shot

curl -X POST https://api.unifically.com/v1/tasks \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "kuaishou/kling-3.0-omni-video",
    "input": {
      "mode": "pro",
      "multi_shots": [
        { "prompt": "Wide shot of city skyline", "duration": 5 },
        { "prompt": "Close-up of glass buildings", "duration": 5 }
      ]
    }
  }'

Response

{
  "code": 200,
  "success": true,
  "data": {
    "task_id": "abc123def456",
    "status": "processing"
  }
}

​Model

​Parameters

​Modes

​Reference images vs elements

​Elements

​Example - Text-to-Video

​Example - With Elements

​Example - Reference Images with Start Frame

​Example - Multi-Shot

​Response

Model

Parameters

Modes

Reference images vs elements

Elements

Example - Text-to-Video

Example - With Elements

Example - Reference Images with Start Frame

Example - Multi-Shot

Response