Clone your voice in Claude and ChatGPT

Yes. Connect Makemancer and Claude, ChatGPT or Codex can clone your own voice, or the voice of someone who has given explicit permission, from one recording of at least 5 seconds; one to two minutes of clean speech works best. Cloning is free and saves the voice on a private Character. Each script it reads costs $0.10 per 1,000 bytes of text, about 1,000 English characters.

Bring
A clean recording of your own voice
Get
A private voice that reads any script
From
10 credits · $0.10

Made with it

Our own results, first takes. Each caption says what it took.

One recording in, new scripts out

For this demo we cloned no one: Maren is a synthetic voice we made with voice design, then recorded reading 53 seconds of plain speech. Cloning that recording took one free request. The clone then read an English intro and a Spanish line that were never in the recording.

InSource recording for the cloning demo: a synthetic voice designed from a text description, not a real person, reading 53 seconds with Fish Drama 3Hi, I'm Maren. This is a short recording for a voice cloning demo, so I'll just talk for about forty-five seconds. I grew up near the coast, and on most mornings I still walk the same two miles along the harbor before work. My favorite stop is the bakery on Fourth Street, which opens at six thirty and sells out of cinnamon rolls by eight. I've tried to bake them at home three times now. The first batch was too dense, the second one burned, and the third was honestly pretty good. Two numbers I always forget: my locker code, which is four, one, nine, seven, and how many plants are on my balcony. I think it's eleven. Anyway, that should be plenty of audio to work with. Thanks for listening.Fish Audio Drama 3 · 7 credits
ResultA clone of a synthetic demo voice, not a real person, reads a new English scriptWelcome back to the studio. Today we're turning one product photo into a fifteen-second ad, a voiceover and a soundtrack, all without leaving the chat. Let's start with the script.Fish Audio Drama 3 · 2 credits
ResultThe same cloned synthetic demo voice, not a real person, reads a line in SpanishHola, soy Maren. Esta misma voz también puede leer tu guion en español, con el mismo tono y el mismo ritmo.Fish Audio Drama 3 · 2 credits

How to do it

  1. Connect Makemancer

    Add Makemancer to Claude, ChatGPT, Claude Code or Codex. The steps for each client are on the connection page.

  2. Record, agree and clone

    Ask to clone your voice. The assistant shows the consent statement to agree to and a private upload for the recording, which never gets a public link. One to two minutes of you talking naturally in a quiet room is ideal. The voice is saved on a Character you name.

    I want to clone my own voice. Tell me what to record and how to upload it privately, show me the consent statement first, and save the voice on a Character called "My voice".
  3. Have it read your scripts

    Name the Character in any request and the script is read in your voice. Directions in square brackets set the delivery; the voice can also read other languages.

    Use my saved voice "My voice" to read this intro, [warm, unhurried]: "Welcome back. Today we turn one photo into a fifteen-second ad." Then read the same intro in Spanish. Tell me the cost first.

Which model to use

ModelBest forPriced forCredits
Fish Audio Drama 3speech/fish-drama-3Speaks every cloned voice, with performance directions in square brackets and other languages from the same voice.per 1,000 UTF-8 bytes10$0.10

Whose voice you can clone

Clone only your own voice, or the voice of someone who has explicitly agreed to be cloned and to have speech generated with it. Before anything is cloned, the assistant quotes a consent statement for you to accept, and that acceptance is recorded with your account and the time. Do not use a cloned voice to impersonate anyone, fake an endorsement or deceive. The full rules are in the voice cloning terms, and what we store and for how long is in the privacy notice.

Record a sample that clones well

  • One voice, one room. No music, no second speaker, no echoing kitchen. A closet full of clothes beats a bare office.
  • Talk, don't perform. Read something casual for a minute or two at your normal pace. The clone copies what it hears, including a flat reading voice.
  • Phone is fine. Hold it a hand's width from your mouth and keep the distance steady.
  • Say numbers and names you use. Product names and figures in the sample help the clone say them the way you do.

Clone a voice or design one

When the voice does not need to be anyone in particular, design one instead: describe it ("warm British woman, relaxed pace, lightly playful") and get three options to audition, at 2 credits a request, no recording or consent needed. The creator on our talking avatar page speaks with a designed voice saved on her Character. Either kind of voice reads scripts the same way, and either can give a talking avatar or a lip-synced clip its words. For stock voices, browse the voice library.

Our demo voice

The voice on this page is synthetic, not a real person. We designed it from a text description, recorded it reading 53 seconds of everyday speech, and cloned that recording, so the demo needed nobody's permission. The source, the English intro and the Spanish line are all above with what each took.

Questions

How do I clone my own voice with AI?

Connect Makemancer, ask the assistant to clone your voice, agree to the consent statement and upload one clean recording. The voice is saved on a Character, and from then on any request that names that Character is read in your voice.

How does AI voice cloning work?

The voice model listens to your recording and learns its timbre, accent and rhythm. It then reads new text in that voice, including words and languages that were never in the recording.

How long does the recording need to be?

At least 5 seconds; one to two minutes of clean solo speech works best, and anything over 2 minutes is cut. Talk naturally, in one quiet room, with no music or other voices.

Whose voice am I allowed to clone?

Your own, or the voice of a speaker who has given you explicit permission to clone it and generate speech with it. Having a recording of someone is not permission, and public figures are not allowed without it. Keep evidence of the permission.

Is the voice private?

Yes. The recording is kept in private storage with no public link and the voice belongs to your account only. Deleting the Character deletes the voice at the provider and the recording from our storage.

What does a cloned voiceover cost?

Cloning is free. Each script the voice reads costs $0.10 per 1,000 bytes of text, about a minute of English speech (Chinese, Japanese and Korean take about three bytes a character). Cloning opens with your first credit pack, and how many voices an account keeps grows with what it has bought.