Kling Motion Control: Copy a Movement onto Any Character

Text prompts are a clumsy way to describe a dance routine, a martial-arts sequence or a precise gesture. Motion Control solves that by letting a video do the describing. You film or choose a movement, give Kling an image of the character who should perform it, and the model generates a new video in which your character moves exactly like the reference. Kling introduced Motion Control with VIDEO 2.6 and upgraded it in VIDEO 3.0, mainly for facial consistency. This guide follows Kling's own user guide and adds the practical choices that decide whether a result is usable.

How it works

Kling describes Motion Control as precise control of a character's movements and facial expressions based on a reference image. Three inputs go in:

1. A motion reference. A video you upload, or a movement picked from the Motion Library.
2. A character image. The person or character who should perform the movement.
3. A prompt. Used to control the background and other scene details, not the movement itself.

The output length matches the usable part of the reference video. Kling's guide notes that if the motion is very complex or fast, the model extracts only the valid, continuous action, so the result can be shorter than your upload, with a minimum of 3 seconds, and credits used in that case are not refunded.

Rules for the reference video

Most failed generations come from the reference video. Kling's requirements and recommendations are specific:

• One person. If several people appear, the one taking up the most of the frame is used.
• One continuous shot. No cuts, shot changes or camera movement; otherwise the video may be truncated.
• The whole body and head visible and not blocked by objects.
• Steady, moderate speed. Very fast motion gives worse results.
• Wide range of motion with little travel. Kling recommends movement that is expressive but stays roughly in place.
• Size and length. The short edge at least 340 pixels, the long edge at most 3,850 pixels, and 3 to 30 seconds long.
• Real human movement works best, although some stylised humanoid proportions are recognised.

Filming your own reference is often better than searching for one: a phone on a tripod, plain background, full body in frame, one take.

Preparing the character image

• Match the framing. Full body in the motion video needs a full-body character image; half body needs half body.
• Leave space for big moves. If the reference has large gestures or steps, give the character room in the frame.
• Clear, unobstructed figure, with the head and limbs visible.
• Proportions close to the reference. Kling advises matching the character's proportions to those in the video.

Character orientation

Kling offers two orientation modes. By default, "Character Orientation Matches Video" makes the character's movements, expressions, camera movement and orientation follow the reference video. "Character Orientation Matches Image" keeps the orientation of your character image and supports camera movement. Use the first when you want a faithful copy of the performance; use the second when your image's composition matters more than the reference's.

VIDEO 3.0 Motion Control and Element Binding

Kling's March 2026 update, VIDEO 3.0 Motion Control, focuses on faces: stable features and smooth expressions across angles, emotions, occlusion and long takes. The tool for that is Element Binding.

1. Upload the reference video and the character image.
2. Click "Bind Facial Element to Enhance Facial Consistency" under the character image.
3. Bind an existing element or create one from a set of images, or by uploading or recording a short video.

The element uses facial information only, not clothing, hairstyle, make-up or props. Kling's advice is to upload face references that match the result you want: front and side views for accurate head turns; neutral and smiling front views for expressions; and a short video for complex emotional changes, because video gives richer facial information. Element Binding only works when the character's orientation matches the video orientation, and only one element is supported per generation. Our guide to Kling Elements covers building elements in general.

Settings at a glance

Credits are charged per second of output, with different rates for Standard and Professional mode; check Kling's current pricing.

Fixing common problems

• The output is shorter than the reference. Parts of the motion were too fast or interrupted. Use a steadier take.
• Limbs bend oddly. The framing of the image and the video differ, or the reference has occlusion. Match full body to full body and keep limbs visible.
• The face drifts. Use VIDEO 3.0 with Element Binding and add front and profile references.
• No element was applied. Several people of similar size were in the first frame, so Kling could not pick one. Crop to one dominant figure.
• The background ignores the prompt. Keep the prompt to scene details; the movement already comes from the video.

Try it

A girl in a white tank top and denim skirt dances on a sunlit rooftop, city skyline behind her, warm late-afternoon light.

Pair it with a 6 to 8 second full-body dance clip filmed on a tripod and a full-body image of your character. Run it once with orientation following the video, then with orientation following the image, and keep whichever composition suits the edit.