How to clone a voice
How to clone a voice, step by step.
A good clone starts with a good recording. This guide shows how to record a clean sample, clone it, and get natural-sounding speech, in a few minutes and without leaving your browser.
How it works
- 1
Make sure you have permission
Clone your own voice, or one you have clear, informed permission to use. You confirm this every time you clone.
- 2
Record a clean sample
10 to 30 seconds of one person speaking naturally, in a quiet room, with no music behind it.
- 3
Upload it and name the voice
Choose the language of the recording, give the voice a name and confirm that you have permission.
- 4
Type your text and generate
Your clone appears with your other voices. Pick it, write your text and generate.
1. Choose a voice you may use
Cloning is powerful, so start with the rule that matters most: only clone a voice that is your own, or when the speaker has clearly agreed to it and to the way you will use the result. Never use a clone to make it seem that a real person said something they did not. Our Terms set out the rules, and so do the conditions of the model we use.
2. Record a good sample
The clone copies everything in the recording, including noise and room echo, so the recording matters more than anything else you do.
- Use a quiet room. Soft furnishings such as curtains and a sofa reduce echo; a bathroom or an empty hall makes it worse.
- Hold the microphone at a steady distance, about a hand’s width from your mouth. A phone voice-memo app is fine.
- Speak naturally, at the pace and in the style you want the clone to have. A flat, bored read gives a flat clone.
- Record one speaker only, with no music, no other voices and no sound effects.
- Aim for 10 to 30 seconds. A longer clean recording is fine: the clearest stretch is picked for you.
3. Clean up a noisy recording
If you could not avoid some noise, run the recording through the free Voice Enhancer first. It removes noise and echo and evens out the volume, and it works in your browser, so the recording is not uploaded.
4. Clone it
Open the Studio, choose Voice Cloning, add your recording, and pick the language it is in. A cloned voice belongs to the language you made it in, and cloning supports English, German, Italian, Portuguese and Spanish. Name the voice, tick the permission box and start. Cloning usually takes more time than using a built-in voice, and how long depends on your device.
5. Get natural results
- Write the way people talk: short sentences and normal punctuation give natural pauses.
- Spell out anything that is read wrongly, such as an unusual acronym or a long number.
- Use the speed and calmness controls to match the mood you want.
- Generate a short test sentence first. If it sounds off, the usual cause is the recording, so try a cleaner one.
If something sounds wrong
- Robotic or buzzy: the recording had noise or echo. Re-record, or clean it first.
- Wrong accent or odd pronunciation: check that you picked the language the recording is in.
- Slow to start: this is normal for cloned voices. For a quick result, use one of the built-in voices in the voice library.
Your recording stays with you.
Cloning happens in your browser. The recording is not uploaded to our servers, and only the saved voice is kept, in your own browser.
Questions
How long should the sample be?
A clear recording of one speaker, ideally 10 to 30 seconds. Longer is fine if it is clean.
Can I clone any voice?
Only your own voice, or a voice whose owner has clearly agreed. You confirm this every time, and misuse can lead to your account being suspended.
Is the clone free?
Yes. A free account can clone a voice and use it within its 5,000 characters a week. Basic, $10 a month or $100 a year, makes it unlimited.
Where is my recording stored?
Nowhere on our side. It is processed in your browser and not uploaded. Only the saved voice is kept, in your browser.
Ready when you are.
Text to speech needs no account to try. A free account gives you 5,000 characters a week and unlimited voice enhancing.
Create a free account