Free Text to Speech — Natural Voice-Overs
Turn text into natural speech with Kokoro, an open Apache-licensed voice model that runs in your browser. Choose from 40 voices in six languages, set the speed and download a WAV or MP3 with no sign-up and no character fees.
Your text stays on your device: the voice is generated in this browser tab.
The voice model is about 90 MB and is downloaded once, then kept by your browser. Languages other than English add about 6 MB.
How it works
Type or paste your script, then pick a language and a voice. The text is split into sentences and spoken by the Kokoro model on your device, and you can listen as soon as it finishes. The first use downloads the model, about 90 MB, and your browser keeps it for next time.
Getting a natural read
Punctuation shapes the delivery, so use commas and full stops where you would pause. Write numbers and unusual names the way they sound if one comes out wrong. A speed between 0.9 and 1.1 usually sounds most natural.
FAQ
Can I use the audio commercially?
Kokoro is released under the Apache 2.0 license, so the audio you make can be used in commercial projects.
Which languages are available?
American and British English, Spanish, French, Italian, Brazilian Portuguese and Hindi. English has the widest choice of voices, and the English voices sound the most natural.
Is there a character limit?
There is no charge per character. Very long texts simply take longer, because everything is generated on your own device.
Is my text sent to a server?
No. The voice is generated in this browser tab.
What can you do with text to speech?
Creators add a voice-over to Shorts and explainers without a microphone; editors drop in a scratch narration before the real recording; course makers and marketers voice training videos and product demos; and writers listen to a draft to catch what reads badly.