Identity across angles
Maintain recognizable facial features during turns, profile views, and changing camera distance.
Character performance, directed by reference
Kling 3.0 Motion Control
Turn one character image and one performance clip into a controllable AI video, with stronger facial consistency through turns, expressions, occlusion, and dynamic framing.
1 image + 1 videoUp to 30 seconds720p or 1080p
Kling 3.0 Motion Control
0/2500 · Optional · up to 2,500 characters
Official product demonstration
See the two inputs and the generated performance together
A street portrait defines the blonde character and background, while the original dance clip supplies the choreography. The output keeps the character's appearance and reproduces the dancer's full-body movement.
Official Kling AI product demonstrationView the official Motion Control guide
Built for expressive character shots
What Kling 3.0 Motion Control improves
The 3.0 model focuses on identity and expression continuity when the performance becomes visually demanding.
Expression and occlusion recovery
Carry subtler expressions and restore facial detail after brief hand or object occlusion.
Dynamic framing and background choice
Handle zooms and orbiting composition, and choose the character image or motion clip as the background source.
From reference motion to character performance
Upload the character
Choose one clear image with a body crop similar to the motion clip.
Add the performance
Upload a single-shot reference video between 3 and 30 seconds.
Direct and generate
Choose orientation and resolution, then use the prompt for scene and camera details.
Choose what controls the character's facing direction
Follow image preserves the reference image orientation and allows up to 10 seconds. Follow video tracks the performer's direction and allows up to 30 seconds.
Follow image
Best when the starting pose and camera direction in the character image should remain dominant.
Follow video
Best for longer performances, turns, and direction changes from the reference clip.
Input requirements
- Use one clear, single-character JPG or PNG image with sides of at least 340 px.
- Use one continuous 3–30 second MP4 or MOV clip with a visible, unobstructed performer.
- Match the framing of the character image and motion video for more stable results.
Need a simpler, lower-cost motion workflow?
Kling 2.6 focuses on full-body motion, gestures, and continuous performances while keeping the same image-plus-video workflow.
Motion Control questions
Practical answers about inputs, duration, quality, and model behavior.