How to swap the person in a video for an AI persona, step by step
You upload a short video, tap the person you want to replace, and an AI persona takes their place in the same scene. Here is the walkthrough in screen order, plus what tends to work and what still goes wrong.

How to swap the person in a video for an AI persona, step by step
You can swap the person in a video for an AI persona in a few taps: upload a 4 to 30 second video, tap the person you want to replace while the video is paused, check the price, generate, and compare the result with the original. The scene, the motion and the audio stay the video's own. Only who appears changes.
I'm the CTO at Cut.Pro and this feature is recent. The walkthrough below follows the screen order, and at the end I say what tends to work and what I would test before betting on it.
What a persona is, in one sentence
A persona is a virtual person whose face, hair, body and style stay the same from video to video. You can pick a ready-made character from the gallery, import your own face with up to 6 photos, import the sheet of an AI influencer you already have, or build one from scratch by choosing traits. This post assumes you already have one. If you don't, start at Studio → AI Personas.
Whether an AI face belongs on your channel at all is something I covered in Digital clones on Shorts: solution or trap?. Here it's only the hands-on part.
Step 1: open the persona and pick the video
Open the persona from the "Your personas" list and click "New video". The form opens with the video type, the upload field and, beside it, the options and the price.
Upload a video of 4 to 30 seconds. The cap depends on the quality you choose:
| Quality | Max length | Accepted file |
|---|---|---|
| Standard | 15 s | MP4 up to 50 MB |
| High fidelity | 30 s | MP4 or MOV up to 100 MB |
Standard is the cheaper option. High fidelity is sharper and closer to the persona. A video slightly over the limit is trimmed to the limit, so a phone clip that measured 15.3 seconds is not a problem. If the length is truly refused, the screen warns you before the click, with the limit of the quality you chose.
My advice for the first video: take a short clip, about 5 to 8 seconds, with one person, well lit, looking more or less at the camera. It's cheap and shows quickly how the persona behaves.
Step 2: choose who becomes the persona
Once the video loads, a "Pick the person" button appears. Pause the video at a moment when the person is clearly visible and tap them. An outline marks who was chosen, and you can tap someone else to switch or clear the choice.
If you pick nobody, Cut.Pro uses the main person in the video. That works fine when there's only one. With two people side by side, choose by hand: it's how you avoid the persona landing on the wrong one. A recent fix helped here, since the persona is reinforced on the chosen person every few seconds, but I still wouldn't skip the pick.
One honest limit: picking by tap works on videos up to 15 seconds. On a longer video you go with the main person.
And "What to replace"?
Just below there's an option for how much changes:
- Head only swaps face and hair;
- Whole person also swaps body and outfit.
If you want to keep the video's clothes and body, stay with "Head only". If the persona has its own look that is part of the point (the outfit, the build), use "Whole person".
Step 3: first-person video? Use "Whoever is filming"
Selfie, daily routine video, unboxing, anything shot with the camera in hand: there's barely a "person" to tap. What you see is hands and, now and then, a face when the camera turns.
For that there's the "Whoever is filming" option, under "Who becomes the persona". It turns the hands, arms and whatever shows of the camera holder's body into the persona, and the face too in the moments the camera turns toward it. It doesn't invent another person in the scene.
An example: an 8-second clip of a hand scrolling a phone and, halfway, the camera flipping to the face of whoever is recording. With "Whoever is filming", the hand and the face both come out as the persona, in the same movement.
I'd stay cautious here: this is where results depend most on the footage. If the camera never shows the face, only hands and arms change. Test with a short clip first.
Step 4: see the price before you generate
The price in credits appears in the form before you confirm, and it changes with the video length and the quality. I'm not putting figures here because they change, and the right number is the one on your screen.
Two things that make testing less scary:
- if the generation fails, the credits come back on their own;
- if you're short on credits, the button takes you to buy them and the generation starts by itself after the purchase, with nothing to redo.
The generate button tells you what's still missing when something is (the video or the person, for example). Once a video is generating, it shows up right away as "generating" in the list, and you can leave the screen. Waiting time varies with video length and how busy things are; I won't promise a number.
Step 5: compare with the original
When the video is ready, it opens in a player built for checking the work, with three views:
- With the persona, the result alone;
- Compare, with a divider you drag ("Drag to compare") to see original and result in the same frame;
- Side by side, both videos in one frame, playing together.
Switching views is instant and the video keeps its place. I use the divider to hunt for flaws, stopping on frames where the person moves fast or turns to profile, and side by side to show the before and after to anyone who doesn't believe it.
Step 6: download, open in the editor, or publish
Like it? Under the player there are three exits:
- Download the file;
- Open in editor, to add captions, trim, adjust the format;
- Publish straight to your connected networks.
The video only shows as ready after the finishing pass that puts the original audio back, so what you download already has sound. Generated videos are also saved in your library.
In Cut.Pro that path is continuous: persona, swap, editor, captions and publishing live in one place, with no exporting and importing files between tools.
What tends to work
- Short video, one person in focus, decent light.
- Medium or close shots, where the face takes a fair share of the frame.
- A scene you already own: the routine shot, the "tutorial in 10 seconds", the reaction.
- A persona made from your own photos, when the idea is to keep your face in videos someone else shot.
What I'd test before betting on it
- Dark, shaky video, or lots of people in the background. I can't guarantee the result. Test with a short clip before spending credits on the long one.
- People crossing in front of each other. Picking by hand helps, but this is the kind of scene where I'd expect more rework.
- Long video. Past 15 seconds you lose the tap-to-pick and depend on the main person.
- Profile faces, a hand crossing the face, fast movement. These are the frames where a flaw shows most. The divider in step 5 exists for that.
A conversation worth having with yourself
Swapping the person in a video for a persona can be just a production tool, or something your audience expects to be told about. Platforms keep changing how they treat AI-made content, and what shifts with content provenance is explained in another post on the blog. My rule: use the AI labels the network offers, and don't hide what you used.
Also remember a persona is your virtual face, not someone else's. Use your own persona or a ready-made character, not the face of someone who didn't agree to it.
Where to start
Tomorrow morning, do just this: record a 6-second video of yourself, open a persona, pick Standard quality, tap yourself with the video paused, check the price and generate. Compare with the divider and decide whether the persona fits your kind of content. If it does, then it's worth thinking about volume, about dubbing into other languages (that's explained in YouTube auto-dubbing in 27 languages) and about where the video will be published.
And if your workflow already goes through an AI editor for Shorts, this breakdown of YouTube's editor helps you decide what goes before the swap and what goes after.


