How Much Audio Do You Need to Clone Your Voice?

For a voice clone you can use in text to speech, a minute or two of clear speech is enough. More audio doesn't help nearly as much as better audio: one quiet, steady minute beats ten noisy ones. The clone copies how you speak as well as how you sound, so record in the style you want it to use.

What makes a good recording?

• A quiet room. A wardrobe full of clothes or a small carpeted room works better than a big empty one.
• A steady distance from the microphone, about a hand's width.
• Only your voice. No music, no other speakers, no notifications. If the recording is noisy, clean it up first.
• The delivery you want. Read naturally, as you would for your videos or podcast. A flat reading makes a flat clone.
• Clean files. In Leelo you can record in the browser, or upload up to five recordings (25 MB in all).

Is it legal to clone a voice?

Your own, yes. Someone else's only with their clear permission to make a synthetic copy. Cloning a celebrity or anyone who hasn't agreed is off-limits on Leelo: you confirm consent each time you clone, and the confirmation is kept with the voice. Voice design likewise refuses voices of children, named people and impersonations.

Can my cloned voice speak other languages?

Yes. A voice of your own can read many languages besides your own. Try a line in each language before a whole project, and use the pronunciation dictionary for names.

What if I don't want to clone a real voice?

You can design one instead: describe the voice you want ("a warm, raspy old sailor who tells stories"), listen to three attempts and keep the one you like. It's a voice that doesn't belong to anyone. See voice design.

What does it cost?

On Leelo, cloning needs the Starter plan or above and costs nothing to make. Speech in a voice of your own costs 5 credits a word, because it's produced by a dearer engine. Each plan includes a number of voices of your own kept for free. See voice cloning and pricing.

What happens to my recording?

It's used to make your voice and is kept with it. Delete the voice and the recording goes too. Leelo doesn't train its own models on your voice or scripts.