Kling V3 Omni

kling-v3-omni is multimodal Kling generation with reference video URLs. Its backend model schema is the source of truth for supported parameters.

Quick Facts

Item	Value
Model ID	`kling-v3-omni`
Type	Video
Region	Global / CN
Best For	multimodal Kling generation with reference video URLs

Recommended Request Shape

Call POST /v1/quote first and send exactly the same parameters when creating the task.

{
  "model": "kling-v3-omni",
  "prompt": "A cinematic product reveal",
  "mode": "pro",
  "audio": false,
  "duration": 8,
  "reference_videos": ["https://cdn.example.com/ref.mp4"],
  "input_video_duration": 8
}

Field Reference

Field	Type	Required	Default	Description
`model`	`string`	Yes	-	Must be `kling-v3-omni`
`prompt`	`string`	Yes	-	Video prompt
`reference_images`	`string[]`	No	-	Reference image URLs
`mode`	`string`	No	`std`	Quality tier: standard, pro, or 4K where supported
`audio`	`boolean`	No	`false`	Whether to generate audio
`duration`	`integer`	No	`5`	Output duration
`aspect_ratio`	`string`	No	`16:9`	Frame ratio
`reference_videos`	`string[]`	No	-	Up to 3 reference video URLs
`input_video_duration`	`number`	With video input	-	Actual processed input-video seconds; must be greater than 0

Supported Enums

Field	Values
`mode`	`std` (720P) `pro` (1080P) `4k`
`audio`	`false` `true`
`duration`	`4` through `15`
`aspect_ratio`	`16:9` `9:16` `1:1`

Kling V3 Omni Integration Notes

When reference videos are present, audio must be false.
has_video_input is derived from reference_videos by the server.
Do not send fields absent from the model schema; use the live GET /v1/models response as authoritative.

Smallest Example

{
  "model": "kling-v3-omni",
  "prompt": "A cinematic product reveal"
}