Home › Photo to Video › Motion Control

How to make a character copy movement from another video

Give the app a picture of your character and a video of the movement you want, and the AI generates your character performing that movement. In the Photo to Video app for iPhone this is Motion Control mode. It runs on Kling 3 at 720p, and it allows the longest clips of any mode here — up to 30 seconds.

Steps

1. Open Motion Control. In the app, go to Modes and choose Motion Control.
2. Add your character. Pick an image of the person or character to animate.
3. Add a reference video. Pick the clip whose movement should be copied.
4. Choose a length. Anything from 2 up to 30 seconds.
5. Generate and save. The character performs the reference movement.

What it supports

Input	One character image plus one reference video
Output	Video, 720p
Length	2–30 seconds — the longest of any mode
Models	Kling 3 (the only model for this mode)
Prompt	Optional
Platform	iPhone (iOS). No Android or web version.
Pricing	Credits per second of output. No subscription.

Which model to use

This mode runs on Kling 3 only, so there is no model choice to make — Kling's handling of human bodies is what makes the transfer hold together. Resolution is fixed at 720p. The variable worth spending attention on is the reference clip, not the model.

Getting a better result

• Pick a reference clip with one clearly visible person, framed head to foot. Crowds and tight crops confuse the transfer.
• Match the framing of your character image to the reference. A close-up portrait driven by a full-body dance rarely works.
• Simple, large movements transfer best. Fast footwork and hand detail are where it breaks down.
• Because clips can run to 30 seconds, cost adds up quickly — prove the pairing at a few seconds first.

What people use it for

Making a character perform a dance or routine

The reference clip supplies timing and body mechanics, which is why this works where a prompt alone would not. Frame your character head to foot to match.

Prompt no prompt needed — the reference video carries the movement

Reusing one performance across several characters

Pick the movement once, then run it against each character image. The motion stays consistent in a way separate generations never would.

Prompt same reference clip, a different character image each run

Giving an illustrated character believable body movement

Drawn and rendered characters work, provided the proportions are roughly human — the transfer maps a skeleton it needs to recognise.

Prompt keep the character fully visible, no tight crop

A longer sequence than other modes allow

This mode reaches 30 seconds, the longest in the app, making it the only route to a full routine in one clip.

Prompt prove the pairing at a few seconds before paying for thirty

Where it struggles

Fast footwork, hand detail and limbs crossing the body are where the transfer visibly fails. Crowded reference clips confuse it — one clearly visible person only. Output is fixed at 720p, so this is not the mode for something that must look sharp full screen.

Questions