AI Motion Control

Put any performance on any character. Add a character image and a motion clip, or pick a move from the library, and Kling 3.0 or Kling 2.6 recreates the body movement, gestures and expressions, with no mocap suit and no keyframes.

  • No mocap suit
  • Kling 3.0 & Kling 2.6
  • Up to 1080p
Keep audio
One reference take, two characters: the same performance transferred onto each of them

Trending dance and motion clips

Popular dances, gestures and action moves, ready to use as your reference. Hover to preview, click to load, then add your character.

15 moves
You provide
1 image + 1 clip
A photo, illustration or 3D render, plus a clip of one person moving
Length
Up to 30 s
The result runs as long as your motion clip
Models
Kling 2.6 or Kling 3.0
720p or 1080p output on either model
Credits
From 14 cr/s
Billed per second of motion clip; exact cost shown before you run

Motion transfer examples

Keep the motion, change the character

Each result shows its two inputs in the corner: the character image decides who performs, the motion clip decides the body movement, gestures and expressions.

Photo to dance video

Turn one photo into a dance video

A full-body photo and a phone-shot dance clip are enough. With Match video, your character steps into the dancer's street and hits every beat, while the face, hair and outfit stay those of the person in your photo.

Try it with your character

Anime and illustration

Give an anime character real gestures

A real performer's heart gestures, hand flicks and head tilts carry over to a hand-drawn character. With Match image, the illustration keeps its own street, line work and colours, and only the performance is borrowed.

Try it with your character

Mascots and non-human characters

Let a mascot borrow a dancer's moves

A crocodile-headed character in streetwear picks up a real dancer's arm swings and bounce without losing its scales, puffer jacket or neon Tokyo street. Brand mascots and game characters get a human performance without any rigging.

Try it with your character

One motion, many characters

Reuse one take across every character

One idol-pose clip drives two different characters, so a campaign gets matching performances without another shoot. Film or pick the move once, then swap the character image for each variant.

Try it with your character

How it works

How do you animate a character with a reference video?

Two inputs and a few choices. No prompt, no keyframes, and nothing is charged until you run.

  1. Add the character image

    A photo, illustration or 3D render of the character you want to see moving.

  2. Add or pick a motion clip

    Upload a clip of one person moving, or pick a move from the motion library. Phone footage works.

  3. Choose what the scene follows

    Match the video or match the image, then pick Kling 2.6 or Kling 3.0 and 720p or 1080p. Keep the clip's audio or switch it off.

  4. Check the cost and run

    The credit cost appears before you confirm. Download the result or send it straight to the upscaler or extender.

Should the scene match the video or the image?

This setting changes the result more than any other. Decide what has to survive: the performer's movement or your image's composition.

  • Match video

    Default
    Keeps
    The performer's turns, facing direction and framing
    Clip length
    Up to 30 seconds
    Best for
    Dances with spins, full routines and longer performances

    Choose it when the movement matters more than the pose in your image.

  • Match image

    Keeps
    Your image's framing and the way the character faces
    Clip length
    Up to 10 seconds
    Best for
    Key visuals, product shots and presenters facing the camera

    Choose it when the picture is already composed the way you want it.

How do you prepare inputs that work?

Most failed runs trace back to the character image or the motion clip, not the model. Check both files against this list before you spend credits.

Inputs that work

  1. The whole body, head to feet

    The model can only move limbs it can see: hands and feet stay in the image, and the performer stays fully in view throughout the clip.

  2. Matching framing in both files

    Full-body photo for a full-body dance, upper-body shot for a seated talk. Mismatched framing stretches or crops the character.

  3. One person, one continuous take

    A single subject in a steady shot, from the first frame to the last, gives the cleanest motion.

  4. Trimmed to the move you need

    The result runs as long as the clip and you pay per second, so cut the idle start and end first.

Inputs that break

  1. Cropped limbs or a hidden face

    Crossed arms, props over the joints and an obstructed face hide the joints and features the motion needs to follow.

  2. More than one person in frame

    Only the largest figure is animated; everyone else is treated as background and can bleed into the movement.

  3. Cuts, transitions or montages

    Every cut is a jump the model tries to animate through.

  4. Occlusion and whip pans

    Walking behind objects, leaving the frame or fast camera swings break the motion.

Use cases

What can you create with motion transfer?

Dance videos are only the start. Motion transfer fits anywhere a real performance needs a different character.

Trend and dance videos

Put a trending dance on your own character for TikTok, Reels and Shorts, with the song still on beat.

Example: Your own photo doing this week's dance challenge for Reels.

Presenters and avatars

Record your own gestures and voice once, and let a virtual presenter or a stylized version of you deliver them.

Example: A 3D avatar opening a tutorial with the host's real hand gestures and voice.

Game and concept characters

See how a character design moves before anyone models, rigs or animates it.

Example: A concept-sheet warrior walking and turning from a reference walk cycle.

Lookbook to runway

Turn a single lookbook photo into a walk or a turn that shows how the outfit moves.

Example: A studio outfit photo animated with a runway turn for a product page.

What are the motion control input specs?

Both tiers accept the same files with the same limits. The result is always exactly as long as the motion clip.

What are the motion control input specs?
Character image
PNG / JPG / WebP · up to 10 MB · one clear figure, full body preferred
Motion clip
MP4 / MOV · 2–30 s · up to 50 MB · up to 60 fps · one person, one continuous take
Orientation
Match the video (up to 30 s) or match the image (up to 10 s)
Output
One MP4 as long as the motion clip; the clip's music and voice carry over by default
Resolution
720p or 1080p output, selectable per run
Pricing
Kling 2.6: 14 cr/s at 720p, 23 cr/s at 1080p; Kling 3.0: 26 cr/s at 720p, 34 cr/s at 1080p. Billed per second of motion clip

Tools for the animated character

Send the performance to the next tool without downloading and re-uploading it.

  • AI Video Generator

    Generate a clip from a prompt, a still image, a reference or existing footage.

    Open tool
  • Video Upscaler

    Raise a clip to 1080p or 4K and sharpen detail frame by frame.

    Open tool
  • Video Extender

    Add seconds before or after a shot while its motion and subject carry on.

    Open tool

Motion control FAQ

What it is, how it differs from image-to-video, length, models, identity, selfies, sound, cost, commercial use and export.

What is AI motion control?+

AI motion control transfers the body movement, hand gestures and facial expressions of a reference video onto a character image, and returns a video of that character performing the same moves. It is also called motion transfer or pose transfer. You need no motion capture suit, no keyframes and no animation skills: one character image and one clip of a person moving are enough.

How is motion control different from image-to-video?+

Image-to-video animates a still from a text prompt, so the model invents the movement. Motion control, also called motion transfer or pose transfer, copies the movement from a real clip, so choreography, timing and gestures come out the way they were performed. Use motion control when you have a reference; use image-to-video in the AI Video Generator when you only have an idea.

How long can the motion clip be?+

Between 2 and 30 seconds when the scene matches the video, and up to 10 seconds when it matches the image. Files can be up to 50 MB. The result is exactly as long as the clip you upload.

What is the difference between Kling 2.6 and Kling 3.0?+

Both take the same image and clip with the same limits, and both render at 720p or 1080p. Kling 3.0 keeps the face and outfit more consistent through fast, complex movement, turns and changing angles, at a higher per-second rate. Start with Kling 2.6, and switch to Kling 3.0 when spins, footwork or hand gestures need more accuracy.

Will the character's face or outfit change?+

The face, hair, outfit and art style come from your image, not from the performer. Drift is most likely when the face is small or hidden, when the character turns fully away, or when the framing of the two files differs. A clear face and matching framing help most, and Kling 3.0 holds up better on fast moves.

Can I animate a photo of myself?+

Yes. Use a clear photo of yourself as the character image and a clip of the moves you want. A full-body photo suits a full-body dance and a waist-up selfie suits gestures or a seated talk; matching the framing of the two files keeps your proportions right.

Can I animate illustrations, 3D characters or animals?+

Illustrations, anime, clay and 3D characters all work when the whole figure is visible and the body reads clearly. The movement comes from a human performer, so characters with a human-like body, a head, two arms and two legs, map best. Four-legged animals and creatures without limbs are much less predictable.

Does the result keep the sound from the motion clip?+

Yes. The clip's audio, whether music, voice or footsteps, is carried into the result by default, so a dance stays on the beat of its song. Switch off Keep audio if you plan to add your own track.

How much does motion control cost?+

You pay per second of motion clip, and 1080p output costs more than 720p: Kling 2.6 at 14/23 credits per second (720p/1080p) and Kling 3.0 at 26/34, rounded up to a whole credit. The exact cost shows before you confirm, and failed runs are refunded automatically.

Can I use motion control videos commercially?+

Subscription plans include a commercial license, and we do not claim ownership of the videos you create. Results made on a free account are public. The rights in your inputs still apply: use a character image and a motion clip you are allowed to use, and get permission for identifiable people, brand characters, and someone else's choreography or footage.

Can I export the motion as FBX, BVH or a rig?+

No. Motion control produces a finished video, not motion-capture data, so there is no skeleton, rig or keyframes to export. If you need animation data for a game engine or 3D software, you still need a mocap or pose-estimation tool.

Motion transfer

Put your next performance on any character

Upload a character image, pick or upload a motion clip, and see the exact cost before you run.