Kling O3 Edit

Kling O3 Edit is the dedicated video-to-video and motion reference framework within Kuaishou's Kling Omni/O3 diffusion architecture. It specializes in instruction-guided video transformation and kinetic camera transfer rather than pure text-to-video synthesis from scratch.

Edit mode directly transforms the visual styling, wardrobe, or environment within your source clip (@Video1) while locking its exact duration, blocking, and camera trajectory. Reference mode extracts only the camera kinetics, pacing, and spatial flow of @Video1 to direct an entirely new synthesized scene with customizable durations from 3 to 15 seconds.

Standard tier renders cost-effective 720p output optimized for rapid stylistic iteration and drafting. Pro tier generates native 1080p resolution with enhanced micro-textures, sharper edge preservation on fast-moving subjects, and refined volumetric lighting for final commercial production.

Tags explicitly instruct the model which media assets govern your transformation. Tagging @Video1 designates the source video clip to edit or borrow kinetics from, while @Image1 through @Image4 anchor uploaded style boards or character reference portraits for identity and wardrobe locking.

The model accepts MP4 or MOV files between 3 and 15 seconds with dimensions from 720 to 3840 pixels per side (up to 200 MB) at 24 to 60 fps. For optimal tracking, supply footage with intentional camera movement and clean subject definition rather than clips with rapid cuts or heavy compression artifacts.