Clone a voice and use it to read your text

Clone a voice and use it to read your text

The job almost always has two halves, and people ask about them together: copy a voice, and have it read a text. The first half happens once and takes twenty seconds. After that the second half is always there: the voice is saved, and it can read as many texts as you like. This page collects everything on the subject: what to record, what you will get, where the limits are, and the line you must not cross.

OnceAs many times as you like10–20 secondsof ordinary speecha phone recorder will doSaved voicestays with you in the chatand does not vanish laterAny textpaste it, get the audioup to 10,000 charactersMost articles explain only the left half, yet people ask about both at once:“clone a voice AND read my text with it”. The right half is the reasonanyone bothers with the left one.
You record once. After that the voice is saved and can read as many texts as you like, with no re-recording on day two or day one hundred.

How it works, step by step

  1. Record 10–20 seconds of ordinary speech. A phone’s voice recorder is enough. Do not read from a page; just tell something in your own words. Natural intonation in the recording becomes natural intonation in the result.
  2. The voice is saved. This is the most underrated part: you do not need to record again tomorrow or a month from now.
  3. Paste a text, get audio. Up to 10,000 characters per request, roughly 10 to 15 minutes of speech. A long text is cut up and joined back automatically; there is nothing to split by hand.
10–20 sOf recording is enoughAnything from 3 to 60 seconds is accepted; more than twenty does not improve the result.
10,000Characters per requestRoughly 10–15 minutes of speech. Split and joined for you.
10Languages supported directlyEnglish, Spanish, Russian, Chinese, German, French and more.

Check your recording first

A clone comes out exactly like the twenty seconds it was made from, and almost everything that can go wrong is visible before you upload.

Is your recording good enough

A clone comes out exactly like the twenty seconds it was made from. Tick everything that is true of your recording and see what you will hear later.

This recording will do

Upload it. The voice will be saved and stay available for any text, so you will not need to record again.

And one rule that has nothing to do with recording quality: you may copy your own voice, or someone else’s with their consent. A recording taken from someone else’s video is still their voice, however easily it uploads.

Try it right here

Below is a working chat, not a picture. Type a text and hear it read aloud; if you want your own voice, send a short recording to the same chat.

Loading…

The line this book does not cross

A voice belongs to a person, not to whoever has the file. You may copy your own voice, or someone else’s with their consent.

A recording cut from someone else’s video, interview or voice message is still their voice, however easily it uploads. There is no technical difference between “mine” and “theirs” here; the whole difference is on your side, and so are the consequences.

Why people do this

  • Your own narration without recording. The script gets edited ten times, and you do not re-record a take each time.
  • One voice across everything. Lessons, how-tos, a podcast, announcements: the same narrator, even on a day they have a cold.
  • Long texts. Reading a fifteen-minute article live is an hour of work with retakes and editing.
  • Fixes after the fact. One figure changed, so you re-voice one paragraph, not the whole video.

Where to go next

The chapters of this book cover particular cases; the full list is below in the contents. If you do not need to copy a voice but do need a text read aloud, see reading a text aloud with a voice. If the text is long and you have hit limits, see long-text narration and its limits. If you need a character’s voice rather than your own, see a character voice for narration. Ethics, risks and how it looks from the side of the person whose voice is copied are in the chapter voice cloning with a neural network.

Frequently asked questions

How do I clone a voice and have it read a text?
In two steps. First send 10–20 seconds of ordinary speech; a phone recorder is enough, and the voice is saved after that. Then paste a text and get audio in that voice, up to 10,000 characters per request. You do not need to record again: the voice stays available for every text that follows. You can try it in the Twelver chat.
Which languages does it work in?
Ten, each supported directly rather than through a translation step: English, Spanish, Russian, Chinese, German, French, Italian, Portuguese, Japanese and Korean. You can record the sample and have the text read in any of them.
How much recording does voice cloning need?
10–20 seconds of ordinary connected speech is the recommended length. Anything from 3 to 60 seconds is accepted, but more than twenty no longer improves the result. What is in the recording matters far more than its length: one speaker, no music or TV in the background, and ideally talking in your own words rather than reading from a page.
Do I have to record my voice again each time?
No. You record once, and the voice is saved and stays available for any later text, tomorrow or a month from now. That is the whole point: twenty seconds once instead of a take for every script.
How long a text can it read?
Up to 10,000 characters per request, roughly 10 to 15 minutes of speech. A long text is split into parts and joined back automatically, so you do not need to cut it by hand. If the text is longer still, split it where the meaning breaks: the voice is kept between requests, and the parts will sound the same.
Can I copy another person’s voice?
Technically the engine does not ask whose voice is in the recording. In substance, a voice belongs to a person, not to whoever has the file: you may copy your own voice, or someone else’s with their consent. A recording cut from someone else’s video or voice message is still their voice, and how easily it uploads does not change that.
What if the clone does not sound like the person?
The cause is almost always the source recording, not the settings. The three most common: music or a TV in the background (the model treats it as part of the voice), two speakers in one file (the voices get averaged), and a recorded phone call (the line cuts what makes a voice recognisable). Re-record twenty seconds in quiet, in your own words; that fixes most cases.

In the Twelver chat you can send twenty seconds of recording and then have that voice read any text: online, in one conversation. Starter tokens arrive on your first visit, and signing up adds more.

Contents

  1. 1.Read a text aloud with a voice: getting speech, not a robot
  2. 2.Long-text narration: where the real limits are
  3. 3.A character voice for narration
  4. 4.Clone a voice free and online

Try it yourself

Everything in this guide runs inside Twelver

One chat for text, images, video, music and voice — no separate services or subscriptions.

Open Twelver chat
Related pageVoice Cloning
Rate your experience