Omni Video

Create a Kling-compatible multimodal Omni video task.

Create Task

Request

POST /v1/videos/omni-video

Field	Type	Required	Supported values and limits	Description
model_name	string	Yes	kling-v3-omni	Omni video generation model. This is the only supported model value at present.
multi_shot	boolean	No	true, false	Enables multi-shot generation. Default is false. When multi_shot=true, prompt is ignored; when multi_shot=false, shot_type and multi_prompt are ignored.
shot_type	string	No	Only customize is supported	Required when multi_shot=true; ignored when multi_shot=false. Only custom storyboards are supported. intelligence is not supported.
prompt	string	No	Up to 2500 chars	Required when multi_shot=false; ignored when multi_shot=true. You can reference input assets in the prompt with <<<image_1>>>, <<<element_1>>>, and <<<video_1>>>.
multi_prompt	array	No	1 to 6 items	Required when multi_shot=true and shot_type=customize; ignored when multi_shot=false. Used to define prompt and duration for each shot.
multi_prompt[].index	integer	Yes	Starts from 1	Shot index.
multi_prompt[].prompt	string	Yes	Up to 512 chars	Prompt for the current shot.
multi_prompt[].duration	string	Yes	At least 1 second	Duration of the current shot. The sum of all shot durations must equal the total duration.
image_list	array	No	Public image URL or raw Base64	Reference image list. It can be used for subject, scene, or style reference, and can also be used as first/end frame input. When an end frame is used, a first frame must also be present; when first-frame or first-and-end-frame generation is used, video editing cannot be used.
image_list[].image_url	string	Yes	.jpg, .jpeg, .png; up to 10MB	Image URL or raw Base64. Images must be at least 300px, and the aspect ratio must stay between 1:2.5 and 2.5:1.
image_list[].type	string	No	first_frame, end_frame	Frame type. If an end frame is used, a first frame must also be provided. End frame only is not supported. When more than 2 reference images are provided, end frame is not supported.
element_list	array	No	Use within scenario limits	Reference element list, suitable for keeping subject identity or visual consistency in the generated video. When first-frame or first-and-end-frame generation is used, up to 3 elements are supported; with a reference video, the combined count of reference images and reference elements must not exceed 4, and video-role elements are not supported; without a reference video, the combined count must not exceed 7.
element_list[].element_id	integer	Yes	Existing element ID	Element ID from the element library.
video_list	array	No	Up to 1 video	Reference video list. It can be used either for video reference or as the input video for editing. When refer_type=base, first/end-frame input cannot be defined; when a reference video is present, sound must be off.
video_list[].video_url	string	Yes	MP4, MOV	Video URL. Input videos must be at least 3 seconds long, no larger than 200MB, within 720px to 2160px in width and height, and between 24fps and 60fps. When using video reference capability, only 3 to 10 seconds is supported.
video_list[].refer_type	string	No	feature, base	feature means reference video. base means the input video will be edited. Default is base.
video_list[].keep_original_sound	string	No	yes, no	Whether to preserve the original input video sound. This also applies to feature reference videos.
sound	string	No	on, off	Whether to generate audio together with the video. Default is off. When video_list is provided, only off is allowed.
mode	string	No	std, pro	Generation mode. std is the standard mode. pro is the higher-quality mode and the default. 4k is not supported.
aspect_ratio	string	No	16:9, 9:16, 1:1	Recommended when the request does not use first-frame input and does not perform video editing.
duration	string	No	3 to 15	Total video duration in seconds. If video_list[].refer_type=base, output duration follows the input video and this field does not take effect.
watermark_info	object	No	{"enabled": boolean}	Whether to return a watermarked result in parallel.
external_task_id	string	No	Custom string	Custom task ID for later query use. It must be unique within the same account.

When video_list is present, set sound to off.

Response

{
  "code": 0,
  "message": "success",
  "request_id": "req_abc123",
  "data": {
    "task_id": "sg_vid_xxxxxxxx",
    "task_status": "submitted",
    "created_at": 1776209922,
    "updated_at": 1776209922
  }
}

Example

Basic Example

curl -sS "https://api.exchangetoken.ai/v1/videos/omni-video" \
  -H "Authorization: Bearer sk-xxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model_name": "kling-v3-omni",
    "prompt": "Make <<<image_1>>> walk through a neon gallery, matching the mood of <<<video_1>>>",
    "image_list": [
      {"image_url": "https://example.com/character.png"}
    ],
    "video_list": [
      {"video_url": "https://example.com/reference.mp4", "refer_type": "feature", "keep_original_sound": "no"}
    ],
    "mode": "pro",
    "sound": "off",
    "aspect_ratio": "16:9",
    "duration": "7"
  }'

First and End Frame Example

curl -sS "https://api.exchangetoken.ai/v1/videos/omni-video" \
  -H "Authorization: Bearer sk-xxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model_name": "kling-v3-omni",
    "prompt": "The subject turns around and the camera slowly pushes in",
    "image_list": [
      {"image_url": "https://example.com/first.png", "type": "first_frame"},
      {"image_url": "https://example.com/end.png", "type": "end_frame"}
    ],
    "mode": "pro",
    "sound": "off",
    "duration": "5"
  }'

Query Task

Request

GET /v1/videos/omni-video/{task_id}

Use data.task_id for polling.

Response

{
  "code": 0,
  "message": "success",
  "request_id": "req_abc123",
  "data": {
    "task_id": "sg_vid_xxxxxxxx",
    "task_status": "succeed",
    "watermark_info": {
      "enabled": false
    },
    "task_result": {
      "videos": [
        {
          "url": "https://example.com/output.mp4"
        }
      ]
    },
    "created_at": 1776209922,
    "updated_at": 1776210072
  }
}

Status Values

task_status	Meaning
submitted	Task accepted but not yet running.
processing	Task is queued or running.
succeed	Task completed and task_result.videos may be present.
failed	Task failed or was cancelled. Check task_status_msg when present.

Example

curl -sS "https://api.exchangetoken.ai/v1/videos/omni-video/sg_vid_xxxxxxxx" \
  -H "Authorization: Bearer sk-xxxxxx"