Skip to main content
Model ID
Calling method: async

Kling V3 Omni & Create Element Integration Guide

This guide covers how to use the Kling V3 Omni video generation model and the Kling Create Element model through the request queue API.

Overview

These two models work together: create elements with kling-create-element, then reference them in kling-v3-omni video generation using element_id and the <<<element_1>>> prompt syntax.

Kling V3 Omni (kling-v3-omni)

Capabilities

  • Text-to-Video: Generate videos purely from text prompts
  • Image-to-Video: Use start frame and/or end frame images to guide generation
  • Video Editing: Edit an existing video (refer_type: "base")
  • Feature Reference: Use a video as a style/motion reference (refer_type: "feature")
  • Element-Driven: Reference custom elements for consistent characters/objects
  • Multi-Shot Storyboards: Create multi-shot videos with per-shot prompts and durations
  • Native Audio: Optional simultaneous audio generation (sound: "on")

Pricing (per second of output video)

When video_list is provided, sound must be "off".

Parameters

Core Parameters

Image & Video Input Parameters (JSON type)

image_list (json, optional)
Reference image list. Images can serve as start/end frames or as element/scene/style references. Types: "first_frame", "end_frame". End frame requires a first frame. Formats: .jpg/.jpeg/.png, max 10MB, min 300px, aspect ratio 1:2.5–2.5:1.
video_list (json, optional)
Reference video. refer_type: "base" (video editing) or "feature" (style/motion reference). keep_original_sound: "yes" or "no". When video_list is provided, sound must be "off". Formats: .mp4/.mov, 3–10s, 720–2160px, 24–60fps, max 200MB.
element_list (json, optional)
Reference elements. Supports three formats that can be mixed in the same list: Mode 1 — Existing elements (by ID):
Mode 2 — Inline image elements (auto-created): Provide frontal_image and refer_images directly. The system auto-creates the element via the Create Element API before generating the video.
Mode 3 — Inline video elements (auto-created): Provide refer_videos directly. The system auto-creates the element via the Create Element API.
Optional fields for inline elements: element_name, element_description, tag_list, element_voice_id. If element_name or element_description are omitted, they are auto-generated. Combined count of images + elements must not exceed 7 (or 4 when video_list is present).

Multi-Shot Parameters

multi_prompt (json, optional)
Per-shot storyboard info. Up to 6 shots, each prompt max 512 chars. Shot durations must sum to total duration. Required when multi_shot is true and shot_type is "customize".

Other Parameters

watermark_info (json, optional)
Whether to generate watermarked results simultaneously.

Limits on Combined References

Submit Request Examples

All examples use POST /requests with the following structure:

1. Text-to-Video (Basic)

Generate a video from a text prompt.

2. Text-to-Video with Native Audio

Generate a video with synchronized audio.

3. Image-to-Video (First Frame Only)

Use a single image as the starting frame.

4. Image-to-Video (First Frame + End Frame)

Guide generation with both a start and end image for smooth transitions.

5. Video Editing (refer_type: “base”)

Edit an existing video. The output duration matches the input video. Note: sound must be "off".

6. Feature Reference Video

Use a video as a style/motion reference (not for editing).

7. Element-Driven Generation

Reference a custom element created via kling-create-element. Use the <<<element_N>>> syntax in the prompt where N corresponds to the order in element_list.

8. Multiple Elements

Reference multiple elements in a single generation.

9. Element + First Frame Image

Combine a custom element with a starting frame image.

10. Element + Feature Reference Video

Combine a custom element with a video style reference. Note: combined images + elements must not exceed 4 when a video is present.

11. Multi-Shot Storyboard

Create a multi-shot video with custom per-shot prompts. The sum of shot durations must equal the total duration.

12. Multi-Shot with Elements

Combine multi-shot storyboarding with custom elements.

13. Inline Image Element (Auto-Created)

Instead of creating elements separately, provide image references directly in element_list. The system auto-creates the element before generating the video.

14. Inline Video Element (Auto-Created)

Provide a video reference directly to auto-create a video character element.

15. Mixed Existing and Inline Elements

Combine pre-created elements (by ID) with inline elements in the same request.

16. Watermark Enabled


Kling Create Element (kling-create-element)

Overview

Create reusable custom elements (characters, animals, items, costumes, scenes, effects) that can be referenced in Kling V3 Omni video generation. Elements maintain consistent appearance across multiple generations. There are two ways to create an element:

Parameters

Core Parameters

Image Reference Parameters (when reference_type is "image_refer")

Video Reference Parameters (when reference_type is "video_refer")

Optional Parameters

tag_list (json, optional)
Tags for the element. One element can have multiple tags.

Available Tags

Submit Request Examples

All examples use POST /requests:

1. Image Reference — Character Element

Create a character element from a frontal photo and side/detail reference images.

2. Image Reference — Animal Element

3. Image Reference — Item/Costume Element

4. Image Reference — Scene Element

5. Video Reference — Character with Voice Customization

Create a character element from video. If the video contains human speech, voice customization is automatically triggered.

6. Video Reference — Character with Existing Voice Binding

Bind a voice from the voice library to an element created from video.

7. Multiple Tags


End-to-End Workflow: Create Element then Generate Video

Step 1: Create a character element

Response:

Step 2: Wait for element creation to complete

Poll GET /requests/abc-123 until status is "success". The outcome will contain the created element with its element_id.

Step 3: Use the element in V3 Omni video generation

Step 4: Retrieve the generated video

Poll GET /requests/{request_id} until status is "success". The outcome will contain: