Generate Videos with Omni

December 11, 2025 (October 20, 2026)

Table of contents

  1. Version Comparison
  2. Request Headers
  3. Request Body
    1. Workflow Selection
    2. Default Workflow (Images/Elements)
    3. Frames Workflow
    4. Video Reference/Transform Workflow
    5. Voices (v4 only)
    6. Duration Support Matrix
    7. Aspect Ratio Support Matrix
    8. Input Limits Matrix
    9. Multi-shot (v3 only)
    10. Other Parameters
  4. Prompt Reference Syntax
  5. Responses
  6. Model
  7. Examples
  8. Try It

This endpoint generates videos using Kling’s Omni model (O1 and v3) and Kling 4.0 (v4). It supports multiple workflows for different use cases:

  • Default workflow: Text-to-video with optional image/element references
  • Frames workflow: Generate video between start and end frame images
  • Video Reference workflow: Use reference videos to guide generation style (up to 5 with v4)
  • Video Transform workflow: Transform/modify a reference video
Version Comparison
Feature v4 v3 O1
Duration 3-30s 3-15s 3-10s
Mode std (720p), pro (1080p) std, pro, 4k std, pro
Multi-shot No Yes No
Native Audio Yes Always on, stereo Yes enable_audio (default on, no video) No
Aspect Ratio 16:9 (default), 9:16, 1:1, 21:9, auto 16:9 (default), 9:16, 1:1 16:9 (default), 9:16, 1:1
Aspect Ratio auto Yes Frames workflow only Frames workflow only
Workflows Default, Video Reference All All
Max image_N 10 7 7
Max element_N 7 7 7
Reference videos Up to 5, 3-30s each, 30s in total 1, 3-15s 1, 3-10s
Voices (voice_N) Up to 3 No No
Prompt length 8000 1700 1700
VIDEO elements Yes Yes No (IMAGE only)

Omni 4.0 costs the same as Kling 4.0 on POST /videos/text2video (see the cost calculator). Kling offers 1080p to paid Kling subscribers only. 4K and HDR are not available for 4.0 yet.

If your Kling account can’t use Kling 4.0 yet, the request returns a 403 error.

https://api.useapi.net/v1/kling/videos/omni

Request Headers 
Authorization: Bearer {API token}
Content-Type: application/json
# Alternatively you can use multipart/form-data
# Content-Type: multipart/form-data
Request Body 
{
  "email": "[email protected]",
  "prompt": "A woman @image_1 dancing elegantly in a ballroom",
  "duration": "5",
  "aspect_ratio": "16:9",
  "count": 1,
  "image_1": "https://s21-kling.klingai.com/ai-platform/xxx/xxx.jpg"
}

  • email is optional. Leave it out and the job goes to one of your accounts that has a free job slot, picked at random. Each account runs at most its maxJobs jobs at once, and GET scheduler/available shows the free slots. If no account has a free slot, the response is 429.
    Pass email to run the job on that account (429 if it is busy). Uploaded files, elements, voices, avatars and earlier tasks belong to one account, so pass the email of the account they came from.
    With more than one account, reading the task with GET tasks/task_id needs its account’s email. The response’s task.userId is that account’s session.userId in GET accounts, and GET scheduler lists each running task with its email.

  • prompt is required (unless using multi-shot), the text description of the video to generate. Maximum length: 8000 characters for v4 and 1700 characters for o1 and v3. Cannot be used together with multi-shot parameters (shot_N_prompt/shot_N_duration). Use @image_1, @image_2, etc. to reference images in your prompt. Use @element_1, @element_2, etc. (or @object_1, etc.) to reference saved elements. Use @video_1 to reference a video in your prompt (@video_1 through @video_5 with v4). Use @voice_1, @voice_2 and @voice_3 to reference voices with v4.

  • omni_version is optional, the Omni model version to use. Supported values: o1, v3 (default), v4. Version v4 uses Kling 4.0. It supports the default and video reference workflows, takes IMAGE and VIDEO elements, up to 5 reference videos and up to 3 voices, always generates stereo audio, and does not take frame_start/frame_end, video_mode: transform or multi-shot. For start, end and keyframe images with Kling 4.0, use POST /videos/image2video-frames with model_name: kling-v4-0.

  • mode is optional, the generation mode. Supported values: std (default), pro, or 4k (4K resolution, omni_version v3 only). Version v4 supports std (720p) and pro (1080p). Kling offers 1080p to paid Kling subscribers only.

  • enable_audio is optional, whether to generate native audio (a soundtrack) for the video. Default: true. Applies to omni_version: v3 without a reference video. Set to false for a silent video, which uses the lower audio-off price tier. Not valid together with video_1 (use keep_audio for video workflows), with omni_version: o1, or with v4, which always generates audio.

Workflow Selection

The workflow is automatically determined based on which inputs you provide:

Inputs Provided Workflow
frame_start (with optional frame_end) Frames
video_1 with video_mode=transform Video Transform
video_1 with video_mode=reference (default) Video Reference
Images/elements only, or text-only Default
Default Workflow (Images/Elements)

Use images and/or saved elements as references in your prompt.

  • image_1 through image_7 are optional, URLs of reference images. You can upload images using POST /assets. Reference in prompt using @image_1, @image_2, etc.

  • image_8 through image_10 are optional, more reference images for v4 only.

  • element_1 through element_7 are optional, IDs of saved elements. Create elements using POST /elements. Reference in prompt using @element_1 or @object_1, etc.

Note: For o1 and v3, images and elements share the same pool of 7 slots. Combined total cannot exceed 7. For v4, you can use up to 10 images and up to 7 elements, with up to 15 references in total, counting image_N, element_N, video_N and voice_N.

Frames Workflow

Generate video that transitions between a start frame and optional end frame.

  • frame_start is required for frames workflow, URL of the start frame image. Internally mapped to image_1.

  • frame_end is optional, URL of the end frame image. Internally mapped to image_2. Cannot be used without frame_start.

Restrictions: When using frames workflow, image_N, element_N, and video_1 are not allowed. Version v4 does not support the frames workflow.

Video Reference/Transform Workflow

Use reference videos to guide generation or transform a video directly.

  • video_1 is optional, URL of a reference video. You can upload videos using POST /assets. Reference in prompt using @video_1. Video must be 3-10 seconds long (O1), 3-15 seconds (v3), or 3-30 seconds (v4).

  • video_2 through video_5 are optional, more reference videos for v4 only. Use them in order starting from video_1, and reference them in the prompt as @video_2 through @video_5. Each video must be 3-30 seconds long, and all reference videos together at most 30 seconds. You can use up to 5 videos and VIDEO elements together, with at most 3 VIDEO elements. Versions o1 and v3 take video_1 only.

  • video_mode is optional, how to use the video reference. Supported values: reference (default), transform.
    • reference: Video guides the style/motion of generated content
    • transform: Video is directly transformed/modified (not supported by v4)
  • keep_audio is optional, whether to keep the soundtrack of video_1. Default: false. Only applies when video_1 is provided. With v4, video_1 must have sound.

Note: When using video_1 with o1 or v3, you can still include up to 4 images/elements (combined). Version v4 keeps its usual limits.

Voices (v4 only)

With v4, people in the video can speak in a voice you choose.

  • voice_1, voice_2 and voice_3 are optional, voices for omni_version: v4. Each takes a voice id or name from GET /videos/voices. To use your own voice, save it first with POST /videos/voices and pass the same email here, because a saved voice belongs to that Kling account. Use them in order starting from voice_1, and reference them in the prompt as @voice_1, @voice_2 and @voice_3, for example A woman says "Good morning" in the voice of @voice_1. Versions o1 and v3 return a 400 error for voice_N.

Some voices have samples shorter than 3 seconds, which Kling does not accept — the request returns a 400 error.

Duration Support Matrix

Duration availability depends on the workflow, inputs, and version:

Scenario O1 Durations v3 Durations v4 Durations Notes
Text-only (no inputs) 5, 10 3-15 3-30 O1 restricted
Single start frame only 5, 10 3-15 No O1 restricted
Images/elements provided 3-10 3-15 3-30 Full range
Both frames (start + end) 3-10 3-15 No Full range
With reference videos Locked to video length Locked to video length 3-30 O1 and v3 cannot be changed, v4 defaults to the total length of the reference videos
  • duration is optional, the duration in seconds. Default: 5. Cannot be used together with multi-shot parameters (shot_N_prompt/shot_N_duration). VIP required for durations 7-10 (O1) or 7-15 (v3). Version v4 takes 3 to 30 seconds. With video_1 and o1 or v3, the duration is the video’s length and cannot be changed.
    With reference videos and v4, you can set 3 to 30 seconds. The default is the total length of the reference videos in whole seconds, from 3 to 30.
Aspect Ratio Support Matrix
Workflow Allowed Values Default
Default (text/images/elements) 16:9, 9:16, 1:1, 21:9 and auto (v4 only) 16:9
Frames workflow auto only auto
Video Reference/Transform 16:9, 9:16, 1:1, 21:9 and auto (v4 only) 16:9
  • aspect_ratio is optional. When using frame_start, aspect ratio is always auto (derived from input). With o1 and v3, auto is NOT allowed for other workflows.
Input Limits Matrix
Workflow Max Images Max Elements Max Combined Max Video Max Voices
Default 7 7 7 0 0
Frames 0 0 0 0 0
Video Reference 4 4 4 1 0
Video Transform 4 4 4 1 0
v4, Default or Video Reference 10 7 15 (video_N and voice_N included) 5 (VIDEO elements included, at most 3 of them) 3
Multi-shot (v3 only)

Split a video into multiple shots, each with its own prompt and duration. Useful for storytelling with scene transitions. Either hand the whole job to the model with multi_shot, or direct each shot yourself with shot_1_prompt and friends.

  • multi_shot is optional, let the model cut the video into several shots on its own. Supported values: false (default) or true. You write one ordinary prompt and Kling decides where the cuts fall. Ignored when shot_1_prompt is present — directing the shots yourself already turns multi-shot on.

  • shot_1_prompt through shot_6_prompt are optional, per-shot text prompts. Maximum length: 2500 characters per shot. Minimum 2 shots required. Cannot be used together with prompt or duration. Use @image_1, @element_1, etc. references within shot prompts.

  • shot_1_duration through shot_6_duration are optional, per-shot duration in seconds. Each shot must have both prompt and duration. Total duration must be 3-15 seconds. Shots must be sequential with no gaps (e.g. shot_1 + shot_2, not shot_1 + shot_3).

Other Parameters

  • count is optional, the number of videos to generate. Range: 1 to 4. Default: 1. VIP required for counts 2-4.

  • maxJobs is optional, range from 1 to 50. Specifies the maximum number of concurrent jobs.

  • replyUrl is optional, a callback URL to receive generation progress and result. See GET /tasks/task_id for response model.

  • replyRef is optional, a reference identifier for the callback.

Prompt Reference Syntax

Reference your inputs in the prompt using these patterns:

Input Type Prompt Syntax Example
Images @image_1, @image_2, … "A woman @image_1 dancing..."
Elements @element_1 or @object_1, … "Character @element_1 walks..."
Frames @image_1, @image_2 "Scene transitions from @image_1 to @image_2"
Video @video_1 (@video_1 through @video_5 with v4) "Scene like @video_1 but in winter"
Voices (v4) @voice_1, @voice_2, @voice_3 'A woman says "Good morning" in the voice of @voice_1'
Responses

  • 200 OK

    {
  "task": {
    "id": 123456789,
    "userId": 12345,
    "type": "m2v_omni_video",
    "scene": "NORMAL_CREATION",
    "status": 5,
    "status_name": "submitted",
    "status_final": false,
    "taskInfo": {
      "type": "m2v_omni_video",
      "inputs": [
        {
          "name": "image_1",
          "inputType": "URL",
          "token": null,
          "blobStorage": null,
          "url": "https://s21-kling.klingai.com/ai-platform/xxx/xxx.jpg",
          "cover": null,
          "fromWorkId": null
        }
      ],
      "arguments": [
        {
          "name": "prompt",
          "value": "A woman Image1 dancing elegantly in a ballroom"
        },
        {
          "name": "rich_prompt",
          "value": "A woman <<<image_1>>> dancing elegantly in a ballroom"
        },
        {
          "name": "kling_version",
          "value": "o1"
        },
        {
          "name": "model_mode",
          "value": "pro"
        },
        {
          "name": "duration",
          "value": "5"
        },
        {
          "name": "aspect_ratio",
          "value": "16:9"
        },
        {
          "name": "imageCount",
          "value": "1"
        }
      ],
      "extraArgs": {},
      "callbackPayloads": [],
      "scene": "NORMAL_CREATION"
    },
    "favored": false,
    "deleted": false,
    "viewed": false,
    "createTime": 1733836800000,
    "updateTime": 1733836800000
  },
  "works": [],
  "status": 5,
  "status_name": "submitted",
  "status_final": false,
  "message": "",
  "limitation": {
    "type": "m2v_omni_video",
    "remaining": 10000,
    "limit": 10000
  },
  "userPoints": {
    "points": [],
    "total": 0
  },
  "userTickets": {
    "ticket": []
  },
  "editProject": null
}

  • 400 Bad Request

    The parameters are invalid. When Kling refuses the submitted parameters, error carries Kling’s own message, for example for a voice whose sample is too short.

    {
  "error": "v40 audio duration must be between 3.0 and 30.1 seconds: audio_1 (TASK.InvalidArguments)"
}

  • 401 Unauthorized

    {
  "error": "Unauthorized",
  "code": 401
}

  • 403 Forbidden

    Your Kling account can’t use the requested model yet, for example Kling 4.0.

    {
  "error": "This Kling account does not have access to the requested model or feature (AUTH.VPrivilegeEscalation)",
  "message": "Service busy (AUTH.VPrivilegeEscalation)"
}

  • 404 Not Found

    {
  "error": "<error message>"
}

  • 500 Internal Server Error

    Kling uses a 500 response to indicate moderation and other issues with the input. It may be hard to separate actual 500 errors from moderation errors, so use the error field text and your best judgement to tell them apart, since the message field most often has very generic and perhaps misleading text.

    {
  "error": "The content you uploaded appears to violate the community guidelines. (CM_EXT.POther)",
  "message": "Service busy (CM_EXT.POther)"
}

When successful, the response includes a task ID which can be used to check the status using GET /tasks/task_id.

Model 
{ // TypeScript, all fields are optional
    task: {
        id: number
        userId: number
        type: string
        scene: string
        status: number
        status_name: 'submitted' | 'failed' | 'processing' | 'succeed'
        status_final: boolean
        taskInfo: {
            type: string
            inputs: Array<{
                name: string
                inputType: string
                token: string | null
                blobStorage: any | null
                url: string
                cover: string | null
                fromWorkId: number | null
            }>
            arguments: Array<{
                name: string
                value: string
            }>
            extraArgs: Record<string, any>
            callbackPayloads: any[]
            scene: string
        }
        favored: boolean
        deleted: boolean
        viewed: boolean
        createTime: number
        updateTime: number
        viewTime: number
    }
    works: Array<any>
    status: number
    status_name: 'submitted' | 'failed' | 'processing' | 'succeed'
    status_final: boolean
    message: string
    error: string
    limitation: {
        type: string
        remaining: number
        limit: number
    }
    userPoints: {
        points: Array<{
            orderId: string
            type: string
            amount: number
            balance: number
            startTime: number
            endTime: number
        }>
        total: number
    }
    userTickets: {
        ticket: Array<{
            orderId: string
            type: string
            packageType: string
            amount: number
            balance: number
            startTime: number
            endTime: number
        }>
    }
    editProject: any | null
}
Examples 
  • # Default workflow with image
curl -X POST "https://api.useapi.net/v1/kling/videos/omni" \
   -H "Content-Type: application/json" \
   -H "Authorization: Bearer ..." \
   -d '{
     "email": "[email protected]",
     "prompt": "A woman @image_1 dancing elegantly in a ballroom",
     "duration": "5",
     "aspect_ratio": "16:9",
     "image_1": "https://s21-kling.klingai.com/ai-platform/xxx/xxx.jpg"
   }'

# Kling 4.0 with images, 21:9
curl -X POST "https://api.useapi.net/v1/kling/videos/omni" \
   -H "Content-Type: application/json" \
   -H "Authorization: Bearer ..." \
   -d '{
     "email": "[email protected]",
     "omni_version": "v4",
     "prompt": "The woman @image_1 walks into the cafe @image_2 and orders a coffee",
     "duration": "12",
     "aspect_ratio": "21:9",
     "image_1": "https://s21-kling.klingai.com/ai-platform/xxx/woman.jpg",
     "image_2": "https://s21-kling.klingai.com/ai-platform/xxx/cafe.jpg"
   }'

# Kling 4.0 with a voice and two reference videos
curl -X POST "https://api.useapi.net/v1/kling/videos/omni" \
   -H "Content-Type: application/json" \
   -H "Authorization: Bearer ..." \
   -d '{
     "email": "[email protected]",
     "omni_version": "v4",
     "prompt": "The dancer from @video_1 performs the moves from @video_2, then says \"Good morning\" in the voice of @voice_1",
     "duration": "10",
     "video_1": "https://s21-kling.klingai.com/ai-platform/xxx/dancer.mp4",
     "video_2": "https://s21-kling.klingai.com/ai-platform/xxx/moves.mp4",
     "voice_1": "Emma"
   }'

# Using saved element
curl -X POST "https://api.useapi.net/v1/kling/videos/omni" \
   -H "Content-Type: application/json" \
   -H "Authorization: Bearer ..." \
   -d '{
     "email": "[email protected]",
     "prompt": "Character @element_1 walking through a garden",
     "duration": "5",
     "element_1": "u_123456789012345"
   }'

# Frames workflow
curl -X POST "https://api.useapi.net/v1/kling/videos/omni" \
   -H "Content-Type: application/json" \
   -H "Authorization: Bearer ..." \
   -d '{
     "email": "[email protected]",
     "prompt": "Smooth transition from @image_1 to @image_2",
     "duration": "5",
     "frame_start": "https://s21-kling.klingai.com/ai-platform/xxx/start.jpg",
     "frame_end": "https://s21-kling.klingai.com/ai-platform/xxx/end.jpg"
   }'

# Video reference workflow
curl -X POST "https://api.useapi.net/v1/kling/videos/omni" \
   -H "Content-Type: application/json" \
   -H "Authorization: Bearer ..." \
   -d '{
     "email": "[email protected]",
     "prompt": "Scene like @video_1 but in a winter setting",
     "video_1": "https://s21-kling.klingai.com/ai-platform/xxx/reference.mp4",
     "video_mode": "reference"
   }'

  • const token = "API token";
const email = "Previously configured account email";
const apiUrl = "https://api.useapi.net/v1/kling/videos/omni";
const response = await fetch(apiUrl, {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    "Authorization": `Bearer ${token}`,
  },
  body: JSON.stringify({
    email: email,
    prompt: "Character @element_1 dancing elegantly",
    duration: "5",
    aspect_ratio: "16:9",
    element_1: "u_123456789012345"
  })
});
const result = await response.json();
console.log("response", {response, result});

  • import requests
token = "API token"
email = "Previously configured account email"
apiUrl = "https://api.useapi.net/v1/kling/videos/omni"
headers = {
    "Content-Type": "application/json",
    "Authorization" : f"Bearer {token}"
}
data = {
    "email": email,
    "prompt": "Character @element_1 walking through garden",
    "duration": "5",
    "aspect_ratio": "16:9",
    "element_1": "u_123456789012345"
}
response = requests.post(apiUrl, headers=headers, json=data)
print(response, response.json())
Try It