Kling 3.0 Omni Edit API

Video to Video

AI video editing with Kling 3.0 Omni. Edit videos using text prompts and reference images.

·Features·FAQ

Documentation

Prompt *

Text prompt describing the edit. Use @image_1, @image_2 etc. to reference images

Upload images/videos above, then type @ in your prompt to reference them

Source video *

Video to reference or transform (output length matches input)

0/1

Click or drag & dropMP4, WEBM, MOV · Max 100MB

Reference images

Up to 4 reference images

0/4

Click or drag & dropPNG, JPG, WEBP, GIF · Max 100MB

Elements JSON

Optional IMAGE/VIDEO elements (see docs). Up to 4; shared pool with image_urls.

Video mode

Reference uses video as style guide; transform modifies the video

Keep audio

Keep original video audio

Mode

Output quality mode

Aspect ratio

Output

Your generated video will appear here

Features

What Kling 3.0 Omni Edit API offers

Video-to-video edits driven by a text prompt and a required source video (output duration follows the input)

Source video up to 10 minutes long per upload constraints in the playground

Up to 4 reference images plus optional IMAGE or VIDEO elements JSON (shared pool of 4 with images)

Reference versus Transform video mode, plus optional keep audio from the source

Standard (720p) or Pro (1080p) output and aspect ratio 16:9, 9:16, or 1:1

REST API with JSON request and response bodies

Use cases

Built for

Primary

Reshoots and revisions - Reprompt the same clip for look, pacing, or subject changes

Reference-guided edits - Tie changes to a small set of reference images or elements

Localization - Adjust visuals while optionally retaining dialogue or ambient sound

Social repurposing - Reframe or restyle clips for different platforms using aspect controls

VFX-style passes - Transform mode when you need a heavier restyle or retime feel

FAQ

About Kling 3.0 Omni Edit API

It takes a source video and a prompt describing the edit. Output length matches the input duration. You can add reference images and optional elements, choose Standard or Pro, and set Reference or Transform behavior for how the model uses the clip.

Use @image_1, @image_2, and so on for reference images. Elements use the structured JSON format from the docs and share the limit with reference images.

The playground accepts source videos up to 10 minutes. Pricing and limits follow your plan and the live docs.

Reference leans on the source clip as a style or motion guide. Transform targets a stronger modification of the supplied footage. Keep audio preserves the original soundtrack when you need sound continuity.