Kuaishou

Kling V3 Omni

VIDEO

Kling v3 Omni video generation model supporting text-to-video, image-to-video, reference-to-video, video-reference generation, and video editing. 4K mode is available for text, image, and reference generation, but not for video-reference generation or video editing.

kling-v3-omni

PROVIDERSKuaishouENDPOINTSvideo-generationCAPABILITIEStext_to_video · image_to_video · video_to_video

PRICING

Pricing is based on generated video duration. Actual charges are based on the final output. When a reference video is used, reference-video pricing applies; otherwise pricing depends on whether audio is generated.

Scenario	Standard	Pro	4K
No reference video, silent	$0.07981 / sec	$0.10642 / sec	$0.39905 / sec
No reference video, with audio	$0.10642 / sec	$0.13302 / sec	$0.39905 / sec
With reference video	$0.11972 / sec	$0.15962 / sec	$0.39905 / sec

Cost: generated video duration × unit price

Parameter	Type	Default	Range / values	Description
`model` *	`string`	`kling-v3-omni`	—	Model name. Use kling-v3-omni.
`prompt` *	`string`	—	1–2500 characters	Video description or editing instruction. Required by this gateway for every Omni task.
`task_type` *	`enum`	`text-to-video`	text-to-video · image-to-video · reference-to-video · video-to-video/reference	Task type. 4k is allowed for text-to-video, image-to-video, and reference-to-video only. video-to-video/reference and video-to-video/edit support standard/std and pro only.
`mode`	`enum`	`pro`	standard · pro · 4k	Quality mode. standard outputs 720p, pro outputs 1080p, and 4k outputs 4K. 4k is supported only for text-to-video, image-to-video, and reference-to-video; it is not supported for video-to-video/reference or video-to-video/edit.
`duration`	`string`	`5`	Task dependent: usually 5 or 10 seconds; video reference supports 3–15 seconds	Output duration in seconds, passed as a JSON string. Text/image/reference generation commonly supports 5 or 10. Video-reference generation supports 3–15. For video editing, output duration follows the input video.
`aspect_ratio`	`enum`	`16:9`	16:9 · 9:16 · 1:1 · auto	Output aspect ratio. auto is supported for video-reference generation; use 16:9, 9:16, or 1:1 for other generation types.
`negative_prompt`	`string`	—	≤ 2500 characters	Content or visual characteristics to avoid. Mainly used for text-to-video.
`cfg_scale`	`number`	—	0–1	Prompt adherence strength for text-to-video, from 0 to 1.
`generate_audio`	`boolean`	`false`	—	Whether to generate synchronized audio with the output video.
`image_url`	`string`	—	—	Primary input image URL for image-to-video.
`start_image_url`	`string`	—	—	Starting-frame image URL for reference-to-video.
`end_image_url`	`string`	—	—	Optional ending-frame image URL. A starting image must also be provided when an ending image is used.
`image_urls`	`array`	—	Up to 7 images; up to 4 when a reference video is present	Reference image URL list for style, subject, scene, video-reference generation, or video editing.
`video_url`	`string`	—	—	Input video URL. Required for video-to-video/reference and video-to-video/edit.
`elements`	`array`	—	Each item uses either element_id or frontal_image_url with optional reference_image_urls	Subject reference list. Each item must either reference an existing element_id or define frontal_image_url with up to three reference_image_urls. Do not mix both forms in one item.
`multi_prompt`	`array`	—	1–6 shots	Multi-shot definitions. Each item may contain index, prompt, and duration. The sum of shot durations must equal the total duration.
`shot_type`	`enum`	`customize`	customize · intelligence	Shot planning mode. customize uses multi_prompt supplied by the client; intelligence lets the model plan shots. Video-reference and video-edit operations support customize only.
`keep_audio`	`boolean`	`true`	—	Whether to preserve the original audio of the input video. Used for video-reference generation and video editing.