API Tool Calls

Free text to speech and voiceover

Type or paste English text, choose a voice and select Speak. Listen as each part finishes, then download your voiceover.

Your privacy: your text never leaves this browser. Speech and audio exports run on your device.

Make your voiceover

0 / 5,000 characters

US English

First Speak downloads a 92 MB voice model, plus about 11 MB runtime, 6–7 MB pronunciation dictionary and a 0.5 MB voice. Cached after the first time when browser storage is available.

Enter English text to start.

WAV: 24 kHz mono. M4A uses AAC when your browser can encode it; otherwise WebM uses Opus.

Words we guessed

Unknown words use a spelling guess. Select a word to add a pronunciation fix.

No guessed words yet.

Pronunciation fixes

Use a familiar English respelling, for example Nguyen = win. Saved on this device when browser storage is available.

AD

Kokoro and Misaki: Apache-2.0 licences and source inventory. MP3 uses LAME / lamejs (LGPL-3.0), served separately and unchanged; complete encoder source. LAME project.

Frequently asked questions

Can I use the voiceover commercially?

Kokoro is licensed under Apache-2.0, which allows commercial use. We give no legal advice; check the rights for your text and intended use.

Which languages and accents are supported?

English only for now, with US and UK voices for women and men. The chosen voice selects the matching English pronunciation dictionary.

Why does the first run download 92 MB?

Natural voices run on the Kokoro model on your device. First Speak downloads 92 MB plus the runtime, chosen voice and dictionary. Browser storage caches them after the first time when available.

How do I fix a mispronounced word?

Choose a word under Words we guessed, or enter any word in Pronunciation fixes. Add an English respelling, such as Nguyen = win, then Speak again. Fixes stay in this browser when storage is available.

More tools

Images & PDFs

Text & codes

Fun & everyday

Money & jobs

All API Tool Calls tools