AI clone video: make an AI video of yourself that sells your product

What an AI clone video is

An AI clone video, also called an AI twin or a digital twin video, is a video of you that you never filmed. A model learns your face from a short recording, then animates it saying whatever script you give it, with the lips moving in time with the words. For a seller, that means the founder or shop owner can introduce a product in every ad without setting up a camera each time.

Three parts make a clone video, and tools differ on each: the face, which comes from a video of you or sometimes a single photo; the voice, which is either a stock AI voice or a clone of yours; and the script, which you write or the tool writes from your product. The more of these is synthetic, the more the video needs a label, and the more it matters that the person in it agreed.

How to make an AI video of yourself

Whatever tool you use, the work is the same six steps. The recording and the check matter most: a clone is only as good as the clip it's made from, and a tool that never checks who is in the clip will also clone someone who never agreed.

1. Step 1

   Record a short clip

   Face the camera in even light, against a plain background, with nothing covering your face: no sunglasses, hat brim or hand. Speak naturally for the length the tool asks for, then hold still with your mouth closed for the last seconds.

2. Step 2

   Prove it's you

   Good tools ask you to read out a code they show only after recording starts, or to turn your head on cue. A video downloaded from someone else's profile can't do that.

3. Step 3

   Wait for the check

   Some tools check the clip automatically, some have a person look at it. Expect anything from minutes to a day.

4. Step 4

   Write the script as the seller

   Hook in the first 3 seconds, the product in the middle, the call to action at the end. Say what the product is, what it does and how to order, in your own words as the person who sells it.

5. Step 5

   Pick the voice

   A stock AI voice is ready at once; a cloned voice needs its own recording and its own consent. Either way, listen to the first take before you pay for more.

6. Step 6

   Label it and post

   Turn on the platform's AI label or add your own, then test the clone against a version with your product alone.

Is an AI clone video a deep fake? The labels that apply

In the EU, it can be. The AI Act defines a deep fake as AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful (Article 3(60)). A realistic video of you saying lines you never filmed fits that description. Article 50(4) makes the business using the AI system disclose that the content is artificially generated, and Article 50(5) wants the disclosure clear and distinguishable at the first exposure. Those duties apply from 2 August 2026, and there is no exception for advertising (EU AI Act).

On TikTok, an ad that is fully AI-generated or significantly edited with AI needs the AI-generated content label or a clear disclaimer, caption, watermark or sticker, and undisclosed AI content can be rejected or restricted (TikTok Ads policy); the built-in disclaimer puts a text label at the bottom of the video (TikTok Ads Help). On YouTube, creators disclose content that is synthetic and realistic enough to be mistaken for a real person or event, with “AI use” under Attributes in YouTube Studio; YouTube's own examples include making it appear someone gave advice they didn't give (YouTube Help). Meta adds an “AI info” label when its own tools add a photorealistic AI person to an ad and says it reads signals such as C2PA metadata from other tools (Meta), but it may not label your clone from them, so say it in the ad text.

New York's 2025 law on synthetic performers works the other way round: it covers digital people who are “not recognizable as any identifiable natural performer” (GBL §396-b), so a recognizable clone of you is outside its definition. That doesn't make the clone unlabelled content anywhere else.

What your AI clone should and shouldn't say

The FTC's Endorsement Guides draw the line that matters. A spokesperson who speaks for the company, describing the product's design, price and how to order, isn't giving an endorsement; someone who says the product is the best they've ever tried is, and an endorser who says they use a product must really have used it (16 CFR Part 255, §255.0 Examples 3 and 6, §255.1(c)). As the seller, your clone is the spokesperson:

• Speak for your shop. “I make these candles in small batches; here's how they're packed” is the seller talking, and it's true.
• Never pose as a buyer. No “I bought this last month”, no reviews, no “verified customer” captions: your clone is the seller, and a fake customer is what the FTC's 2024 rule on reviews and testimonials targets.
• Only results you can prove. A clone saying “it cleared my skin” makes the same claim you would, with the same need for evidence.
• Only your own face. A clone of a creator, an employee or a celebrity needs their permission; the FTC says a celebrity's avatar used without it would violate its rule (FTC), and TikTok removes ads that use a private person's likeness without permission.

An AI clone video generator for product ads: ProductToReel

ProductToReel, which we make, turns a product link or photos into 12–20-second vertical video ads, and since October 2026 the presenter in them can be an AI avatar of your own face. You set it up once after sign-in: you sign a release, then record a 8–20-second clip on your phone, reading aloud a code that appears only once you're recording and turning your head. A person at ProductToReel checks the clip, usually within a day, and a 5-second hello video shows you your avatar. The setup costs 2 credits, which come back if your face is never verified.

Then pick “You” as the presenter of Presenter + product, where you say the hook and the call to action around clips of your product, for 2 credits a video, or Product in hand, where your avatar holds the product, for 3 credits in two steps. You can also choose one of 16 synthetic presenters instead. Every video with your face shows “AI avatar of the seller” for its whole length, speaks for your shop as the seller in a stock AI voice rather than yours, and carries C2PA Content Credentials, though platforms may not label your ad from them. It's one face per account, your own, 18 or older and not a public figure, and you can delete it at any time. Someone else's face, a cloned voice and faceless Hands only videos with your face aren't offered.