Kling o3 4K Reference Video to Video generates a new video at the 4K resolution guided by a reference video plus any images and elements you supply. Unlike the edit tool, this is a fresh generation rather than a change applied to your footage.
In your prompt, use @Video1 for the reference video, @Image1 and @Image2 for reference images, and @Element1 and @Element2 for elements.
Elements are characters or objects you want to include in the result. Supply a clear front view of each one, plus other views to define its appearance from different angles. When a video is supplied, reference images and elements are limited to four in total.
Reference video requirements: .mp4 or .mov, 3 to 15 seconds, 720 to 3840 pixels, at most 200MB. Aspect Ratio set to Auto follows the reference video. Video Duration runs from 3 to 15 seconds and cost scales with it.
Example: a new scene from a reference video
Reference video: a tracking shot across a museum forecourt.
Kling o3 4K reference-video result: a new canal-side university scene guided by the source camera movement and a text prompt.
Pick a model and start generating
These models are tuned for detail — open one and put this guide to work.
