How to Make an AI Clone of Yourself

What exactly is an AI clone?

Your clone is two trained models working together: a likeness model that reproduces how you look and move on camera, and a voice model that reproduces how you sound. Feed it text and it returns a finished talking video of you. Nothing about your face or your voice is invented — it is rebuilt from the samples you supplied, which is why sample quality decides everything.

How do you record footage your clone can actually use?

1. 01Light your face, not the roomOne soft source in front of you, window or panel. Side shadows confuse the likeness model and produce that rubbery look people blame on the software.
2. 02Shoot on a plain, still backgroundA blank wall beats a busy office. Movement behind you gets learned as part of you and shows up as artifacts in every render.
3. 03Talk like a person for a full minuteRead something conversational with normal hand movement and blinks. Frozen posture trains a frozen clone.
4. 04Record voice separately, in a quiet spaceA closet or a car is fine. Keep the room noise out; the voice model copies hiss just as faithfully as it copies you.
5. 05Render a test script before you commitHave the clone read 20 seconds of your real ad copy. Judge it on lip sync and pacing, not on the first two words.

How do you clone your voice so it still sounds like you?

Voice models copy energy, not just timbre. If you read your sample like a bedtime story, every ad your clone films will sound sleepy. Record the sample at the volume and speed you would actually use selling something you love, include a laugh and a question, and the model keeps that range available for later scripts.

• Read one paragraph flat and one paragraph excited so the model learns both ends.
• Say your own brand names and product names out loud in the sample — otherwise the model guesses pronunciation.
• Never mix two recording spaces in one sample; the model averages them and sounds unstable.
• Re-record the sample any time your on-camera style genuinely changes.

What should the clone say?

The clone is an actor, not a strategist. It performs whatever structure you hand it, so the script carries all the weight: a hook in the first two seconds, one specific problem, one demonstration or joke, one offer, one instruction. Kristine's ad scripts run this shape whether a human or a clone reads them.

Hook

The first sentence, written to stop a thumb. Concrete, strange, or funny — never a greeting.

Problem beat

One recognizable frustration, described the way your buyer describes it.

Proof beat

A number, a screen recording, or a visual gag that makes the claim credible.

Ask

A single instruction. Two asks in one ad cuts response roughly in half.

Where do people go wrong with AI clones?

• Training on old, poorly lit footage and blaming the tool for the result.
• Writing 90-second monologues — clones expose padding faster than a live read does.
• Publishing one clone video instead of ten variations, which wastes the whole advantage.
• Hiding that a clone was used in contexts where disclosure matters, such as testimonials.
• Cloning yourself before you have an offer worth filming.

Can you get paid to build clones for other people?

Yes, and it is the fastest freelance angle in this space right now. Businesses want the output but will not learn the workflow. Members of Kristine's community charge per finished video for exactly this — record the client once, keep their clone on file, then deliver batches of ads on request.