Text to Speech
Turn text into natural-sounding speech for free — choose from 28 American and British English AI voices, adjust the speed, and download the result as an MP3 or WAV file. Speech is generated on your own device by the open-source Kokoro voice model, so your text is never uploaded, there's no sign-up and no character quota beyond 20,000 per run.
100% private — everything runs in your browser. Conversion happens on your device; nothing is uploaded.
The first time, the voice model (about 90 MB) downloads once and is then cached by your browser.
Open-source credits: voice model Kokoro-82M by hexgrad (Apache-2.0); kokoro-js (Apache-2.0) and Transformers.js (Apache-2.0); pronunciation by phonemizer.js, which includes eSpeak NG — licensed under the GNU GPL v3, loaded unmodified, with full source code available at those links; MP3 encoding by lamejs (LGPL).
Speech is generated on your device by an open-source AI voice model — your text is never uploaded. English only (American and British voices). Very long texts are read sentence by sentence; on a phone, shorter texts work best.
How to convert text to speech
- Type or paste your text.
- Choose a voice and speed.
- Click Generate speech (the voice model downloads once the first time).
- Play it, then download the MP3 or WAV.
What you can use it for
- Create voice-overs for videos, reels and presentations
- Listen to articles, notes and drafts while you work
- Make audio versions of blog posts and newsletters
- Prototype narration for e-learning and product demos
More free tools
Designing with text?
Style, caption and decorate text with our free, private text and image tools.
See all free tools →Frequently asked questions
Is this text to speech really free?
Yes. There's no sign-up, watermark or paid tier — the voice model runs on your own device, so it costs us nothing per use.
Can I use the audio commercially?
The Kokoro voice model is released under the Apache-2.0 licence, which allows commercial use. As always, make sure you have the rights to the text you're reading.
Why does the first run take a while?
The first time, your browser downloads the voice model (about 90 MB). After that it's cached, and generation starts straight away.
Which languages are supported?
English — 20 American and 8 British voices. Other languages aren't supported by this voice model yet.
Is my text uploaded?
No. The text is turned into speech entirely in your browser; only the voice model itself is downloaded.
Should I download MP3 or WAV?
Choose MP3 for most uses: it's a small file that plays almost everywhere, encoded here at 128 kbps mono. Choose WAV if you plan to edit the audio further in a video or audio editor, because it's uncompressed 16-bit mono. Both files contain exactly the same speech, and both are made and saved straight from your browser.
Is there a character limit?
Each run reads up to 20,000 characters, and a counter under the text box shows how many you've used. There's no daily quota or account, so for a longer script, split it into parts and generate each one in turn. Long text is read in groups of whole sentences, with a short pause added between paragraphs, so leave a blank line wherever you want a break.
Can I change the reading speed?
Yes. Use the Speed slider before you click Generate speech: it goes from 0.5× (half speed) to 2× (double speed) in small steps, with 1× as the normal pace. The speed is built into the audio itself, so the MP3 or WAV you download plays at the speed you chose. Slower settings can help with learning and pronunciation, faster ones with long drafts.
What does the 'Use my GPU' option do?
It runs the voice model on your graphics card through WebGPU, which is usually faster, but the one-time download is about 310 MB instead of about 90 MB. The option only appears in browsers that support WebGPU. If speech generation fails with it ticked, untick it and try again: the standard mode runs on your processor and works in more browsers.
Which voice sounds the most natural?
Heart and Bella, both American English female voices, are marked as best quality in the voice list, so they're a good place to start. There are 28 voices in all, 20 American and 8 British, male and female. Voices differ in tone and pacing, so try a short sentence with two or three of them before you generate a long script.
What's the difference between text to speech and speech to text?
Text to speech turns written words into spoken audio, which is what this tool does. Speech to text works the other way round, turning a recording or your voice into written words. For that, use our free Voice to Text tool for live dictation, or Audio to Text to transcribe a recording you already have.