Kling 2.6 Motion Control
What is Kling 2.6 Motion Control?
Kling 2.6 Motion Control transfers motion from a reference video onto a character image. You upload a clip of someone dancing, walking, or moving plus a still of your character (or mascot, or actor), and the model returns a new video where your character performs the reference motion. Two modes drive the result: Video orientation for body motion (dance, full-body action, articulated movement) and Image orientation for camera-style moves (where the framing changes more than the body). Reference clips up to 30 seconds in Video orientation, capped at 10 seconds when Image orientation is selected. Output is Standard 720p or Pro 1080p.
Key features of Kling 2.6 Motion Control
Four features define what production teams use the endpoint for.
Motion transfer from a reference video
Pass a short motion clip and a single character image. The model produces a new video where the character performs the motion in the reference. No rigging, no full motion capture, no per-frame keyframing.
Two character-orientation modes
Video orientation follows the motion reference and is tuned for full-body articulation up to 30 seconds. Image orientation preserves the character pose for camera-driven shots, capped at 10 seconds.
Standard or Pro output
Std mode renders at 720p for iteration and short-form social work. Pro mode renders at 1080p for delivery cuts where resolution is the contract. Same input pair, same prompt, different render.
Audio passthrough on by default
keep_audio is on by default, so the soundtrack of the reference video carries into the output. Useful for music-driven dance transfers and lip-aligned performance swaps; turn it off when you plan to add a fresh score.
Best for
Trend choreography swaps
Take a viral dance, apply it to your character. Useful for brand participation in social trends without filming the choreography.
Animated mascot performances
Drive a brand still or 2D mascot through real human reference motion. Useful for ads and brand activations.
Fitness and coaching content
Show form once on a template clip, apply it to multiple presenter images for a series.
Indie game character motion
Rapid character motion previews without a full rig. Useful for early-stage game prototyping.
Camera-style moves with Image orientation
When the framing should change but the character should not articulate much (slow zooms, dollies, pans).
Audio-preserved performance clips
Keep the original soundtrack in place when the audio is the point of the clip.
Variants
Two character-orientation modes plus two output tracks. Combine them based on what the reference clip does and where the result will run.
Video orientation
Use when the motion is in the body: dance, walking, full-body action, articulated movement. Uses the full 30-second reference range. Default for trend choreography and mascot performance work.
Image orientation
Use when the motion is camera-driven: slow zooms, dollies, pans, framing changes. Caps usable motion at 10 seconds, and the character articulates less. Good when the lens does the work.
Std
The 720p output mode. Faster turnaround and lower cost per call.
Pro
The 1080p output mode. Same motion-transfer model, higher resolution. Use it when the cut goes to a paid placement or a hero unit.
Use cases
Build a brand mascot performance from a single piece of artwork by recording a 10-second motion reference on a phone and passing both to the endpoint with character orientation set to video. Run a fitness series with one trainer image by templating the motion clip and looping the same character image with new prompts. Convert a dance trend into a brand activation by handing the original choreography reference and your own character still to Pro mode, with audio passthrough on so the music carries through. Shoot a slow-zoom hero moment without a real camera move by switching to Image orientation and feeding the model a 10-second pan reference.
API examples
Call Kling 2.6 Motion Control from any language by POSTing to /v1/tasks. Full parameter docs live at docs.unifically.com/models/video/kling/kling-2.6-motion-control.
curl -X POST https://api.unifically.com/v1/tasks \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "kuaishou/kling-2.6-motion-control",
"input": {
"prompt": "Person dancing gracefully",
"video_url": "https://example.com/dance-motion.mp4",
"image_url": "https://example.com/person.jpg",
"character_orientation": "video",
"mode": "pro",
"keep_audio": true
}
}'
Successful submission returns a task_id. Poll GET /v1/tasks/<task_id> or set a callback_url on the request to receive the finished result.
FAQs
People also ask
It returns a video where a character image performs the motion from a supplied reference clip. Useful for trend choreography, animated mascots, fitness coaching, and indie game character work without rigging or full motion capture.
Reference clips up to 30 seconds. Video orientation uses the full 30-second range for body motion. Image orientation caps usable motion at 10 seconds and is tuned for camera-style moves rather than full body articulation.
It tells the model whose orientation should win. Set it to video for dance, walking, and full-body action where the motion reference should drive the result. Set it to image for camera-driven shots where the framing changes more than the body.
Motion Control 3.0 tightens facial consistency through fast moves and adds inline elements for props or partner characters. 2.6 Motion Control matches existing pipelines that already use these timings and assets. If face stability is the bar, use 3.0; if pipeline parity matters more, stay on 2.6.
Yes. keep_audio defaults to true, so the soundtrack of the reference video carries into the output. Turn it off when you plan to add a fresh score or VO.
Three required inputs: a short prompt describing the motion, the subject character as an image_url, and the motion reference as a video_url. The character image must show the subject's head, shoulders, and torso clearly.