Kling 3.0 Omni Edit
What is Kling 3.0 Omni Edit?
Kling 3.0 Omni Edit is the video-to-video editing route under Kling 3.0 Omni. Pass a source video (up to 10 minutes from the playground) and a prompt describing the edit, and the model returns a new clip of the same length. The reference bundle is up to 4 across image_urls and elements combined (shared cap of 4). Two video modes drive the result: Reference uses the source as a style or motion guide while the prompt steers the change; Transform reshapes the source more aggressively. keep_audio preserves the original soundtrack, which is useful when audio continuity matters. Output renders at Standard (720p) or Pro (1080p), in 16:9, 9:16, or 1:1.
Key features of Kling 3.0 Omni Edit
Five features cover what teams build with Omni Edit.
Edit a source video with a single prompt
Pass video_url and a prompt describing the edit. Output length follows input length, so the timeline stays put while the visual changes.
Reference and Transform video modes
Reference treats the source as a style or motion guide for the prompted change. Transform reshapes the source itself with a heavier restyle or retime. Pick the right one for how aggressive the edit needs to be.
Image and video elements in the reference bundle
The elements array supports both image and video element types. Useful when the reference is motion or style from another clip rather than a still. Combined cap of 4 with image_urls.
Audio passthrough on demand
Set keep_audio to true to preserve the source soundtrack in the output. Default is off, which is what you want when you plan to score the edit in post.
Standard or Pro output
Std mode renders at 720p for editorial passes and alternate-cut review. Pro mode renders at 1080p for delivery. Three aspect ratios cover most placements.
Best for
Reshoots and revisions
Reprompt the same clip for look, pacing, or subject changes without rebuilding the timeline.
Reference-guided edits with VIDEO elements
Use video elements as conditioning. Useful when the reference is motion or style from another clip, not a still.
Localization with audio preserved
Adjust visuals while keeping the dialogue track. Useful for ads going to multiple regions with the same VO.
Social repurposing
Reframe and restyle clips for different platforms using aspect controls and prompt adjustments.
VFX-style passes via Transform
Heavier restyle or retime when the brief needs more than a colour tweak.
Long-source workflows up to 10 minutes
Most edit endpoints cap at 30 seconds. Omni Edit handles a 10-minute source, which supports trailer cuts and long-form work.
Variants
Two output tracks and two edit modes. Combine them based on resolution and how aggressive the edit needs to be.
Standard
The 720p output mode. Faster turnaround for editorial passes where the cut is not heading to a hero unit yet.
Pro
The 1080p output mode. Use it when the edit is the delivery cut and resolution is the contract.
Reference mode
Treats the source clip as a style or motion guide for the prompted change. Good for edits that should preserve the original feel while the prompt steers a specific change: costume swap, lighting shift, character substitution.
Transform mode
Reshapes the supplied footage more aggressively. Heavier restyle, retime, or scene-level rework. Use it when the edit is a meaningful departure from the source.
Use cases
Re-cut a finished spot for a new region by passing the original ad as video_url, prompting in the new background or signage swap, and keeping audio passthrough on so the VO stays locked. Restyle a brand piece into a different look by switching to Transform mode and passing a reference image of the target style as @image_1. Build alternate cuts for client review without re-shooting by reprompting the same source three different ways and sending the chosen variant on Pro. Apply a VFX-style colour and motion pass to a trailer in one call instead of layering multiple passes in an NLE.
API examples
Call Kling 3.0 Omni Edit from any language by POSTing to /v1/tasks. Full parameter docs live at docs.unifically.com/models/video/kling/kling-3.0-omni-video-edit.
curl -X POST https://api.unifically.com/v1/tasks \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "kuaishou/kling-3.0-omni-video-edit",
"input": {
"prompt": "Transform into anime style",
"video_url": "https://example.com/video.mp4",
"video_mode": "reference",
"mode": "pro"
}
}'
Successful submission returns a task_id. Poll GET /v1/tasks/<task_id> or set a callback_url on the request to receive the finished result.
FAQs
People also ask
It is the video-to-video editing route under Kling 3.0 Omni. You supply a source video and a prompt describing the edit. Output length matches input. Up to 4 total across image_urls and elements (shared combined cap of 4) condition the change. Pick Standard or Pro and Reference or Transform behavior.
Source uploads up to 10 minutes per the playground constraint. Output duration is locked to the input video length and cannot be set manually.
Reference treats the source clip as a style or motion guide for the prompted change. Transform reshapes the supplied footage more aggressively (heavier restyle or retime). keep_audio preserves the original soundtrack when sound continuity matters.
Both reshape an input clip with the same Reference and Transform mode choice. Omni Edit allows VIDEO elements alongside image references; O1 Edit drops the elements parameter and only takes image_urls. Pick Omni Edit when the brief needs a video element in the reference bundle.
Use @image_1, @image_2, etc., for reference images, and named element tags (@element_1, @element_2) for entries in the elements array.
16:9, 9:16, or 1:1. The output preserves the chosen aspect.
Modèles liés
Voir tous les modèlesKling 3.0 Omni
Up to 7 references plus elements, video reference/transform, multi-shot 2–6, 4K mode.
- Text to Video
- Image to Video
- Reference to Video
- Video to Video
Kling O1
Reasoning video gen with up to 7 references, image-only elements, single-shot 3–10s, 720p/1080p.
- Text to Video
- Image to Video
- Reference to Video
- Video to Video