Kling Motion Control 3.0
What is Kling Motion Control 3.0?
Kling Motion Control 3.0 is Kuaishou's motion-transfer video model. Pass a character image, a short prompt, and a motion reference video, and the model returns a clip of that character performing the motion. The 3.0 generation tightens facial consistency over the 2.6 release, so the subject's face stays recognisable through head turns, lighting changes, and aggressive choreography. The endpoint is single-purpose by design: there is no text-only mode. Output inherits aspect ratio from the character image and length from the motion reference clip, so the shot is planned up front by choosing the right two assets. Output renders at Std (720p) or Pro (1080p), with audio passthrough on by default.
Key features of Kling Motion Control 3.0
Five features define what production teams use this model for.
Facial consistency under fast motion
The 3.0 generation tightens the model around the subject's face. Head turns, lighting changes, and aggressive choreography no longer drift into a different identity halfway through. The cleaner your character portrait, the stronger this effect lands.
Two output resolutions, one decision
Std mode renders at 720p and is good for fast iteration, mood checks, and short-form social work. Pro mode renders at 1080p and is the delivery resolution. Pick the resolution at the end of the cycle.
Pose vs motion priority you can choose
The character_orientation setting lets you decide whose pose wins. Follow the motion reference for dance and full-body articulation. Preserve the character image's pose for camera-driven shots, lens motion, and stylised loops where the body should stay close to the original artwork.
Audio passthrough on by default
The reference soundtrack carries through to the output unless you turn it off. That keeps music timing and dialogue locked to the motion, which is what you want for dance transfers, lip-aligned performance swaps, and music-driven mascot work.
Best for
Brand mascot animation
Front-facing mascot artwork plus a captured motion reference yields a clip of the mascot performing on-brand. Facial structure stays stable across fast moves.
Choreography and dance transfer
Hand the model a dance reference and a character still. Motion follows the video, the face follows the still, and audio passthrough carries the music through.
Recurring social personas
A single hero character can perform new content week after week from the same reference still. Useful for serial creators and short-form skit channels.
Lip-aligned performance swaps
Set character orientation to video and keep audio enabled to swap a recorded actor for an illustrated character without redoing the audio.
Educational presenter explainers
Show a gesture pattern once, then apply it to a presenter image to reuse the same motion across multiple lessons.
Variants
Motion Control 3.0 has two output resolutions and two pose-priority modes. Combine them based on whether motion or character pose should win, and where the cut will run.
Std
The 720p output mode. Good for fast iteration, mood checks, and short-form social work.
Pro
The 1080p output mode. Same motion-transfer model, same input pair. Use it for delivery cuts where 1080p is the contract.
Video orientation
Set character_orientation to video when motion should win. Good for dance, full-body articulation, and choreography transfer where the reference's movement should come through cleanly.
Image orientation
Set character_orientation to image when the character's pose should be preserved. Better for camera-driven shots, lens motion, and stylised loops where the body should hew close to the original artwork.
Use cases
Animate a brand mascot from a single piece of static artwork by recording a 10-second movement reference on a phone and giving the model both. Build a weekly serial around an illustrated character without redrawing each frame. The same character image plus new motion clips becomes the production pipeline. Swap an actor for an animated character in an existing skit by sending the original clip as the motion reference and a character still. Audio passthrough keeps the dialogue and music intact, so post is just a colour pass.
API examples
Call Kling Motion Control 3.0 from any language by POSTing to /v1/tasks. Full parameter docs live at docs.unifically.com/models/video/kling/kling-3.0-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-3.0-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
Kling Motion Control 3.0 is Kuaishou's motion-transfer video model. Hand it a character image and a motion reference video and it returns a clip of that character performing the motion. The 3.0 generation is tuned for stronger facial consistency, so the subject's face stays recognisable through fast articulation.
Three. A short text prompt describing the motion, a character image showing the subject, and a video providing the motion to transfer. The character image needs the head, shoulders, and torso clearly visible so the model can anchor the upper body during the transfer.
Resolution. Std mode outputs 720p, Pro mode outputs 1080p. Use std for fast iteration and short-form social work, and pro for delivery cuts where 1080p is the contract.
Whose orientation wins when motion and pose disagree. Set it to video to follow the motion reference, which is good for full-body articulation like dance. Set it to image to preserve the input character's pose, which is better for camera-driven shots and lens motion.
Yes. Audio passthrough is on by default, so the reference soundtrack is carried into the output. Turn it off when you plan to add a fresh score or VO. Useful for music-driven dance transfers and lip-aligned performance swaps.
A single front-facing or three-quarter image with the subject's head, shoulders, and torso clearly visible and not occluded. The clearer the upper-body anchor, the better Motion Control 3.0 holds facial structure during fast or complex motion.