Kling Omni Video

Channel type: Kling Omni Video (63) Gateway path: POST https://maasunion.com/kling/v1/omni/videos · GET https://maasunion.com/kling/v1/omni/videos/{task_id} Official upstream: https://api-beijing.klingai.com (/v1/videos/omni-video)

Overview

The Omni models can combine a Prompt with images, subjects, videos, and other content to achieve multiple capabilities, including:

• Text-to-Video, Image-to-Video, first & last frame to video
• Image/subject reference to video
• Instruction transform (video editing)
• Video reference (camera motion/style/generate next shot)
• Multi-shot video (kling-v3-omni)
• Sound generation (kling-v3-omni, sound: on)

Reference materials in the prompt using <<<>>>, for example:

• <<<image_1>>> — the 1st reference image (corresponds to image_list[0])
• <<<element_1>>> — the 1st subject (corresponds to element_list[0])
• <<<video_1>>> — the 1st reference video (corresponds to video_list[0])

This gateway keeps the request/response format identical to the official Kling OmniVideo API; the request body is passed through to the upstream as-is.

Supported Models

Authentication

The upstream channel side uses official Kling authentication, handled automatically by the gateway; the client does not need to and should not pass the upstream key.

Prompt Reference Syntax

• The prompt must not exceed 2,500 characters
• Specify subjects, images, or videos using the <<<>>> format:
  • <<<image_1>>>, <<<image_2>>> — reference the images at the corresponding index in image_list
  • <<<element_1>>>, <<<element_2>>> — reference the subjects at the corresponding index in element_list
  • <<<video_1>>> — reference the video at the corresponding index in video_list
• Different model versions and video modes support different ranges; see the Kling Capability Map

Create Task

POST https://maasunion.com/kling/v1/omni/videos

Submit a video generation task. On success it returns a task ID, which you can use to poll the result via the Query Task endpoint.

Request Body Parameters

multi_prompt Description

Define the index, prompt, and duration of each shot via index, prompt, and duration:

• At most 6 shots, at least 1
• Each shot prompt ≤ 512 characters
• Each shot duration ≥ 1 second and ≤ the total task duration
• The sum of all shot durations equals the total task duration

"multi_prompt": [
  { "index": 1, "prompt": "string", "duration": "5" },
  { "index": 2, "prompt": "string", "duration": "5" }
]

image_list Description

Reference image list, including subjects, scenes, style reference images, etc.; can also be used as the first or last frame to generate video.

Image requirements:

Other rules:

• If an image is not the first or last frame, do not configure the type parameter
• Last frame only is not supported (when there is a last frame image, a first frame image must be present)
• When generating video from a first frame or first & last frame, the video editing feature cannot be used
• When using kling-video-o1, setting first/last frames is not supported when there are more than 2 images in the array
• The image_url parameter value must not be empty

element_list Description

Subject reference list, configured based on the IDs of subjects in the Kling subject library.

Subjects are divided into video custom subjects (video character subjects) and image custom subjects (multi-image subjects), with different applicable scopes.

video_list Description

Reference videos, obtained via URL. Can be used as a feature reference video or as a video to be edited.

Video requirements:

Other rules:

• When the reference video is a video to be edited (refer_type=base), you cannot define video first/last frames
• When there is a reference video, the sound parameter value can only be off
• The video_url parameter value must not be empty

mode Description

Request Example

Image-to-Video

curl --request POST \
  --url "https://maasunion.com/kling/v1/omni/videos" \
  --header "Authorization: Bearer sk-xxxxxxxx" \
  --header "Content-Type: application/json" \
  --data '{
    "model_name": "kling-video-o1",
    "prompt": "让<<<image_1>>>中的人物向镜头挥手",
    "image_list": [
      { "image_url": "https://example.com/reference.png" }
    ],
    "duration": "5",
    "mode": "pro",
    "aspect_ratio": "16:9"
  }'

Multi-Shot (kling-v3-omni)

curl --request POST \
  --url "https://maasunion.com/kling/v1/omni/videos" \
  --header "Authorization: Bearer sk-xxxxxxxx" \
  --header "Content-Type: application/json" \
  --data '{
    "model_name": "kling-v3-omni",
    "multi_shot": true,
    "shot_type": "customize",
    "prompt": "",
    "multi_prompt": [
      { "index": 1, "prompt": "<<<image_1>>>A person sitting on a park bench, sunlight filtering through trees.", "duration": "2" },
      { "index": 2, "prompt": "A car speeding down a rainy street, headlights glowing.", "duration": "3" }
    ],
    "image_list": [{ "image_url": "https://example.com/frame.png" }],
    "mode": "pro",
    "sound": "on",
    "aspect_ratio": "16:9",
    "duration": "5"
  }'

Response Example (200)

{
  "code": 0,
  "message": "string",
  "request_id": "string",
  "data": {
    "task_id": "task_xxxxxxxx",
    "task_info": { "external_task_id": "string" },
    "task_status": "submitted",
    "created_at": 1722769557708,
    "updated_at": 1722769557708
  }
}

Note: The response structure is identical to the official Kling one; data.task_id is the public task ID assigned by the gateway. Please save this ID for querying.

Query Task

GET https://maasunion.com/kling/v1/omni/videos/{task_id}

Request Example

curl --request GET \
  --url "https://maasunion.com/kling/v1/omni/videos/your-task-id" \
  --header "Authorization: Bearer sk-xxxxxxxx"

Response Example (Success)

{
  "code": 0,
  "message": "string",
  "request_id": "string",
  "data": {
    "task_id": "task_xxxxxxxx",
    "task_status": "succeed",
    "task_status_msg": "string",
    "watermark_info": { "enabled": false },
    "task_result": {
      "videos": [
        {
          "id": "string",
          "url": "string",
          "watermark_url": "string",
          "duration": "5"
        }
      ]
    },
    "task_info": { "external_task_id": "string" },
    "final_unit_deduction": "string",
    "created_at": 1722769557708,
    "updated_at": 1722769557708
  }
}

Scenario Examples

Image/Subject Reference

curl --location "https://maasunion.com/kling/v1/omni/videos" \
  --header "Authorization: Bearer sk-xxxxxxxx" \
  --header "Content-Type: application/json" \
  --data '{
    "model_name": "kling-video-o1",
    "prompt": "<<<image_1>>>在东京的街头漫步，偶遇<<<element_1>>>和<<<element_2>>>，并跳到<<<element_2>>>的怀里。视频画面风格与<<<image_2>>>相同",
    "image_list": [
      { "image_url": "https://example.com/a.png" },
      { "image_url": "https://example.com/b.png" }
    ],
    "element_list": [
      { "element_id": 123456789 },
      { "element_id": 987654321 }
    ],
    "mode": "pro",
    "aspect_ratio": "1:1",
    "duration": "7"
  }'

Instruction Transform (Video Editing)

curl --location "https://maasunion.com/kling/v1/omni/videos" \
  --header "Authorization: Bearer sk-xxxxxxxx" \
  --header "Content-Type: application/json" \
  --data '{
    "model_name": "kling-video-o1",
    "prompt": "给<<<video_1>>>中的穿蓝衣服的女孩，戴上<<<image_1>>>中的王冠",
    "image_list": [{ "image_url": "https://example.com/crown.png" }],
    "video_list": [{
      "video_url": "https://example.com/input.mp4",
      "refer_type": "base",
      "keep_original_sound": "yes"
    }],
    "mode": "pro"
  }'

Video Reference (Camera Motion/Style)

curl --location "https://maasunion.com/kling/v1/omni/videos" \
  --header "Authorization: Bearer sk-xxxxxxxx" \
  --header "Content-Type: application/json" \
  --data '{
    "model_name": "kling-video-o1",
    "prompt": "参考<<<video_1>>>的运镜方式，生成一段视频：<<<element_1>>>和<<<element_2>>>在东京街头漫步",
    "element_list": [
      { "element_id": 123456789 },
      { "element_id": 987654321 }
    ],
    "video_list": [{
      "video_url": "https://example.com/ref.mp4",
      "refer_type": "feature",
      "keep_original_sound": "yes"
    }],
    "mode": "pro",
    "aspect_ratio": "1:1",
    "duration": "7"
  }'

First & Last Frame

curl --location "https://maasunion.com/kling/v1/omni/videos" \
  --header "Authorization: Bearer sk-xxxxxxxx" \
  --header "Content-Type: application/json" \
  --data '{
    "model_name": "kling-video-o1",
    "prompt": "视频中的人跳舞",
    "image_list": [
      { "image_url": "https://example.com/start.png", "type": "first_frame" },
      { "image_url": "https://example.com/end.png", "type": "end_frame" }
    ],
    "mode": "pro"
  }'

Capability Matrix

Element Count Limits

Without Reference Video

With Reference Video

• Video character subjects are not supported
• Total reference images + multi-image subjects ≤ 4

First & Last Frame Scenarios

• When kling-v3-omni uses first & last frames, it supports up to 3 subjects
• When kling-video-o1 uses first & last frames, subjects are not supported

FAQ

When can duration be specified?

• Text-to-Video, Image-to-Video (without first & last frame): 3~10 seconds optional
• Video editing (refer_type=base): cannot be specified; aligned with the input video
• Other scenarios (no video + images/subjects, or video + refer_type=feature): 3~10 seconds optional

When is aspect_ratio required?

• Not supported: Instruction transform (video editing), Image-to-Video (with first & last frame)
• Usually required: Text-to-Video, image/subject reference, video reference, etc.

Relationship between multi_shot and prompt?

• When multi_shot=true, prompt is invalid; you must use multi_prompt (shot_type=customize) or prompt (shot_type=intelligence)
• When multi_shot=false, shot_type and multi_prompt are invalid

Notes

1. Asynchronous task: The create endpoint only returns a task ID; you must poll the Query Task endpoint to get the video URL.
2. Video URL validity: Valid for about 30 days after generation (anti-leech format); please persist in time.
3. model_name is required: The request body must include model_name.
4. Base64 images: The gateway automatically removes the data:image/...;base64, prefix before forwarding to the upstream.
5. With a reference video: sound can only be off.
6. Gateway query: Use the gateway task_id from the create response to query; do not use the original upstream ID.

For more general information, see Kling Gateway Overview.