Exact choreography
Arm swings, turns, steps and their timing are copied from the reference video, beat for beat.
Model
Type
Prompt
0/2000
Try an example
CostFree
Four photos, moved by a reference video
Kling 3.0 Motion Control
Photo: a man in a navy kurta in a garden. Reference: the hoodie dance clip. Orientation: follow video.
Kling 3.0 Motion Control
Photo: a woman in a mustard saree in a garden. Reference: the hoodie dance clip. Orientation: keep photo.
Kling 3.0 Motion Control
Photo: a woman in a black tank top in a bedroom. Reference: a 5 second clip of a woman walking on a beach. Orientation: follow video.
Kling 3.0 Motion Control
Photo: a young man in a white t-shirt on a quiet street. Reference: the hoodie dance clip. Resolution: 1080p.
1Upload a photo
One person with head, shoulders and torso in view.
2Add a reference video
A 3 to 30 second MP4 or MOV of the movement you want.
3Pick resolution and run
720p costs 10 credits and 1080p 13 credits per second of video.
Your generations will show up here.
Kling 3.0 Motion Control: your photo, someone else's moves
Kling 3.0 Motion Control takes the person in one photo and makes them perform the body movement, gestures and expressions from a reference video. The face, outfit and setting come from the photo; the timing and choreography come from the video. On Buble it renders 720p or 1080p clips as long as the reference, from 3 to 30 seconds.
Four photos, moved by a reference video
Every clip was generated on Buble for this page from one still photo and one reference video, listed under each example. The people are fictional, generated with AI, so no real person's photo or footage was used. Each clip kept the face, clothing and background of its photo and followed the timing of its reference, including a full turn away from the camera and back.
Inputs
Photo: a man in a navy kurta in a garden. Reference: the hoodie dance clip. Orientation: follow video.
The dance reference behind three of the examples
This four second clip of a dancer in a grey hoodie drove the kurta, saree and street examples above. It was itself generated on Buble, so no real dancer's performance was copied. Any clear video of one person moving works the same way: a phone recording, a stock clip you have the rights to, or a clip generated with another model.
How the clip was made
Full-body shot of a dancer in a grey hoodie doing a simple hip-hop groove with arm swings and a small spin, static camera, plain studio wall
How it works
From one photo and one video to a finished clip
Motion control splits a video into two parts that you supply separately: who appears on screen, and how they move. That is why the same reference can be reused for many different people, and the same photo can be given many different routines.
Step 01
The photo sets the look
Face, hairstyle, clothing, body proportions and the background all come from the still image you upload.
Step 02
The video sets the motion
Body pose, steps, arm and hand movement, head turns and expressions are read from the reference clip frame by frame.
Step 03
Orientation decides the direction
Follow video turns the person the way the performer faces; Keep photo holds the direction of the photo for shorter clips.
Step 04
The render matches the reference length
A 4 second reference gives a 4 second clip, so you control the duration and the price by trimming the reference first.
What it does
What Kling 3.0 Motion Control gives you
Describing a dance in words rarely lands; showing it does. Motion control replaces the prompt with a performance.
Your look stays
The face, hair, clothing and background of the photo carry through, as in all four examples above.
Gestures and expressions
Hand movements, head turns and facial expressions are transferred along with the body.
Two orientation modes
Follow the dancer's direction, or keep the way the person faces in the photo.
Up to 30 seconds
The clip is as long as your reference, so longer routines are possible in one render.
720p or 1080p
Draft in 720p, then render the keeper in 1080p for a sharper final.
Specs
Parameters and input limits
Everything you can set for Kling 3.0 Motion Control on Buble, and what each input accepts.
Character photo
One JPG or PNG up to 10 MB, at least 340 px on each side, aspect ratio between 2:5 and 5:2, with head, shoulders and torso visible.
Reference video
One MP4 or MOV, 3 to 30 seconds long and up to 100 MB. The output has the same length.
Prompt
Optional, up to 2,500 characters, for scene, lighting or style. The movement always comes from the video.
Resolution
720p or 1080p. 720p is the default.
Orientation
Follow video (default, references up to 30 s) or Keep photo (references up to 10 s).
Price
10 credits per second at 720p and 13 at 1080p, counted on the reference video and rounded up to the next second.
Use cases
What people make with it
Anything where a specific movement matters more than a written description.
Dance and trend videos
Put yourself or a character into a choreography you have on video.
Character animation
Make an illustrated or 3D character perform an actor's movements.
Product and fashion clips
Show an outfit on a model photo with a catwalk turn or a natural walk.
Previsualisation
Block out a scene with a rough phone video, then render it with the final look.
Tips
How to get a clean transfer
The clearer both inputs are, the closer the result follows them.
Match the framing
A full-body reference works best with a full-body photo; a waist-up photo gives a waist-up video.
One person in each input
Use a reference video with a single performer and a photo with a single person.
Steady camera in the reference
A static or smoothly moving camera makes the motion easier to follow.
Use footage you may use
Upload your own videos or ones you have the rights to, and photos of yourself or people who agreed.
FAQ
Kling 3.0 Motion Control FAQ
Answers about inputs, length and credits.
Create
Try Kling 3.0 Motion Control
Upload a photo and a short reference clip; a 5 second 720p render costs 50 credits.