01
Upload one clean motion reference
Use a continuous 3–30 second clip with one clear subject. Avoid cuts, abrupt scene changes, and motion that leaves the frame.
Upload a character image and a motion reference video. Kling 3.0 transfers the action while preserving the subject, expression, and performance rhythm.
It is a reference-led video workflow: the image defines who appears, while the reference video defines how that subject moves.
Use a clean character image for appearance and a single continuous clip for motion. Kling 3.0 maps the action and expression rhythm onto the new subject.
Read the official Kling AI Motion Control guideThe model separates appearance from action, then gives you one orientation choice that changes how the movement is interpreted.
Two clear inputs
Reference image · Motion video · Optional prompt
Identity stability
Angle changes · Head turns · Dynamic framing
Expression fidelity
Expressions · Occlusion · Facial clarity
One decisive setting
Image orientation · Video orientation
These limits reflect the current EvoLink API documentation and the controls available in the KlingVideo generator.
Required inputs
Reference duration
Output quality
Image upload
Video upload
Prompt
Character orientation
Reference sound
Choose the source that should control body direction. This choice changes the maximum duration and the kind of movement the model handles best.
| Parameter | Matches image | Matches video |
|---|---|---|
| Maximum duration | 3–10 seconds; maximum 10 seconds. | 3–30 seconds; maximum 30 seconds. |
| Body direction | Follows the pose and orientation in the character image. | Follows the posture and direction in the motion reference. |
| Best fit | Shots that need more camera movement around the character. | Complex actions where matching the reference movement matters most. |
| Upstream Element Binding | Not compatible. | Required upstream, but Element Binding is not exposed in the current KlingVideo generator. |
Give appearance and movement separate jobs, then choose the orientation that matches the shot you need.
01
Use a continuous 3–30 second clip with one clear subject. Avoid cuts, abrupt scene changes, and motion that leaves the frame.
02
Match full-body or half-body framing to the motion clip, and leave enough space around the subject for the action.
03
Use image orientation for camera movement or video orientation for complex action, then generate at 720p or 1080p.
It transfers the movement trajectory from a reference video to the character or object shown in a reference image.
Upload one directly accessible character image and one directly accessible motion reference video. A text prompt is optional.
The accepted range is 3–30 seconds. Image orientation supports up to 10 seconds, while video orientation supports up to 30 seconds.
Match the image when camera movement matters more. Match the video when the action is complex and the body direction should follow the reference clip.
Yes. The current integration offers 720p and 1080p output quality.
The current KlingVideo integration keeps the original reference-video sound by default.
Use one clear subject, continuous movement, moderate speed, and stable framing. Avoid cuts, fast motion, and scene changes.
No. Kling documents Element Binding for its upstream video-orientation workflow, but the current KlingVideo generator does not expose that control.
Commercial use depends on your active plan and the rights attached to the uploaded image, reference video, likeness, audio, and brand assets.
Start with one character image and one clean motion reference, then generate a performance you can judge and refine.