Voice Cloning: Clone Your Own Voice with AI

What is voice cloning, and what makes a good clone?

Voice cloning is AI that learns the sound of one speaker from a short recording: the tone, the accent, the rhythm and the small habits that make a voice recognisable. After that, the clone reads any script you type in the same voice, so you write a line instead of recording it.

A clone is as faithful as its sample. A clear, natural recording gives the AI an honest picture of your voice, and the result sounds like you on a good day. That is why the next steps start with the sample.

How to record a great voice cloning sample

The clone copies what it hears, so the room matters as much as the voice. Record in a quiet room with soft surfaces, turn off fans and music, and keep the microphone a hand's width from your mouth for the whole take.

Read in the tone you want back: calm for lessons, upbeat for a lively show. Thirty seconds to a minute of clean speech is the sweet spot. Uploading a file? Pick a clean recording of just you speaking.

1. 1Pick the quietest room you have; curtains, a sofa and a rug soak up echo.
2. 2Keep the microphone a hand’s width from your mouth, at the same distance for the whole take.
3. 3Read at your natural pace in the tone you want back, and let small pauses happen.
4. 4Keep the live meter steady in the middle, and take a sip of water before you start.

Which microphone should you use?

A studio is optional. A recent phone held at chest height, or a headset with a boom microphone, records clean speech for a clone, and the browser checks the level as you go.

A USB microphone helps most in a slightly noisy room, because it picks up your voice more than the space around it. Whatever you choose, use the same one for every take; the clone learns the sound of the microphone as well as your voice.

Record a ten second test, play it back on headphones and listen for hum, echo or clipped peaks. A quick test is the easiest way to a clean clone.

Write scripts that sound like you

A cloned voice sounds most like you when the script sounds like you, too. Use the words and phrases you say every day, keep sentences short, and write numbers and names the way you pronounce them.

Tags add the moments a plain script misses: [pause] before a key point, [laughs] after a light line, [sighs] for a relaxed aside. Add a word to your pronunciation list once, and every script reads it your way.

Clone my voice for other languages

Once you clone your voice, it reads 36 languages in text to speech. Paste a translated script, keep the same voice, and viewers in other markets hear the presenter they already know.

Record your sample in the language you speak most naturally; your tone and rhythm carry across. For names and brand words in a new language, check the first lines and add any fixes to your pronunciation list.

Voice cloning consent and the law

Whether a clone is lawful depends above all on consent. Tennessee's ELVIS Act protects a person's voice, including a simulated one, from unauthorized use, and New York's right of publicity law requires written consent to use a voice for advertising or trade. Get the owner's written permission before you clone a voice that isn't yours.

The EU AI Act asks anyone publishing AI audio that resembles a real person to disclose it; a note such as "narrated with an AI version of my voice" is enough.

Get the most from your cloned voice

Open text to speech, pick your voice under My voices and paste a script. Short sentences and everyday words sound most like you.

Clone when listeners should hear you. For a character or a second speaker, design a brand new voice in the style you like instead, or pick one of the 300+ preset voices included in text to speech; designed and cloned voices sit side by side in My voices.