Skip to main content
Motion Control takes motion from a reference video and applies it to a still image. The subject in your image performs the same movements as the subject in the reference video. This gives you precise control over animation without needing to describe motion in words.
Diagram showing a reference dance video on the left, a still character image in the middle, and the output video on the right where the character performs the dance

Motion Control concept: a reference video's motion applied to a still image.

​
How motion control works

1

Prepare your inputs

You need two things:
  • A still image - The subject you want to animate (a person, character, mascot, or product).
  • A reference video - The motion you want to apply (a dance, a walk, a gesture, or any movement).
2

Provide both to the agent

Upload both the image and the reference video using the + button. Describe what you want in the prompt.

Make a character perform a dance from a reference.

Animate a product mascot with a wave gesture.

3

Wait for generation

The AI processes the motion reference and applies it to your image. A placeholder appears on the canvas while it generates.
4

Review the result

The output video appears on the canvas. Your subject performs the motion from the reference video while keeping its original appearance.
Canvas showing the output video of a character performing the dance from the reference video

The output video on the canvas showing the animated character.

Motion Control creates a new canvas item. The original image and reference video are preserved. The output video keeps the original sound from the reference video by default.

​
When to use this

  • Making a character or mascot perform a specific dance, gesture, or action.
  • Animating a product with precise, repeatable motion.
  • Creating consistent motion across multiple characters using the same reference video.
  • Producing social media content where a specific movement is required.

​
Character orientation

The agent decides whether the character’s body follows the reference video’s orientation or stays in the original image’s pose. By default, the body orientation follows the reference video for the most natural result.

​
What you cannot do

  • You cannot use motion control with a video as the subject. The subject must be a still image.
  • You cannot combine motion from multiple reference videos in a single generation.
  • You cannot selectively apply motion to parts of the subject (for example, only the arms).
  • You cannot control the duration independently. The output matches the reference video length.

​
Next steps

Edit a Video

Restyle or modify the output video after motion control.

Combine into Timeline

Add the motion-controlled clip to a multi-clip sequence.