The output follows the reference action and its overall timing closely.
Recreate · 4 min read
How to recreate a video with your own character
Bring a character you own — a mascot, an AI character, an illustration — and have it perform an existing clip instead of shooting anything.
Recreate with your characterUpdated 2026-09-04
The short answer
A recreation needs two things: a reference performance and a character. The reference drives motion, camera, timing and audio; your character image decides who is on screen. That split is why you never write a prompt — you are not describing a video, you are casting one. Pick a template or upload a clip you own, add one image of your character, and generate.
Step by step
- Understand the two inputs
The reference clip is the performance — motion, framing and timing are driven by it. Your character image is the casting. Because the motion is not invented from a text description, the result tends to look consistent rather than dreamlike.
- Choose a reference that suits your character
A whole-body action like a dunk or a runway walk needs a character with a readable silhouette. A seated speaker suits a head-and-shoulders image. Matching the two is the difference between a clean transfer and a stylised one.
- Prepare the character image
One subject, front-facing, evenly lit, nothing covering the face. Illustrated and 3D characters work as well as photos, as long as the subject is clearly separable from the background. You must own or be authorised to use the character.
- Generate, then reuse the same character
Once a character works well, run it through other templates. The character image is the reusable asset — the same photo can go on a stage, a runway and a red carpet, which is far cheaper than staging any of them.
What the reference controls
Framing and camera movement are driven by the reference clip, not re-invented.
The output duration follows the reference duration, and key beats are designed to stay aligned.
With “Keep the original sound” selected, the reference audio is carried into the result.
Burned-in captions and stickers are generally preserved with the reference, but generative output can vary — review the final result before publishing.
UI, product shots and price badges are designed to stay in place. Small text and fine detail are worth checking on the result.
What actually makes a difference
A four-legged pet on a human walk cycle reads as deliberate and stylised, not broken — but know that going in rather than being surprised by it.
A busy background makes your character harder to separate. A plain wall or a clean studio shot gives the cleanest transfer.
Every second costs credits. Prove the character works on a 6-second reference before committing to a 15-second one.
Mascots, brand characters and AI characters are fine if they're yours. Do not cast a real person you have no permission to depict, and do not imply an endorsement.
Questions
Is there a prompt box?
No. The reference clip is the instruction. That is a deliberate design choice — describing motion in text is exactly where video models are least reliable.
Can I use an illustrated or 3D character?
Yes. Illustrations, mascots and 3D renders all work, provided the subject is clearly separable from the background.
Can I use my own footage as the reference?
Yes. Switch to “Upload your own video” and bring an MP4 or MOV between 3 and 15 seconds, up to 50MB, that you have the rights to use.
How many templates can one character image be used in?
As many as you like. The image is reusable; each generation is billed separately by output duration.
Related tutorials
Ready to try it?
Bring a character you own — a mascot, an AI character, an illustration — and have it perform an existing clip instead of shooting anything.
Recreate with your character