At a glance
- Accepts
- Typed or pasted text, or a .txt or .md file
- Outputs
- Live playback and a 24 kHz mono download as MP3 (64 or 128 kbps, encoded in the browser) or 16-bit WAV
- Where it runs
- Entirely in your browser, on your device
- What leaves your device
- Nothing - your text never leaves your device; the voice model is downloaded to you
- Limits
- English only (American and British voices); first use downloads a 92 MB model (CPU) or 326 MB (GPU); speed depends on the device, and the finished audio is held in memory at about 2.9 MB per minute
- Account
- Not required - there is no sign-up
- Price
- Free, with no watermark and no usage cap
- Made by
- Visrut
- Last updated
- 12 October 2026
Paste any amount of text, pick one of 28 English voices, and press Read aloud. Playback starts as soon as the first sentence is ready, and the rest is read in the background while you listen. When it finishes you can download the whole thing as an MP3 or a WAV file. It is free, there is no character limit, and your text never leaves your device.
How it works
The voices come from Kokoro-82M, an open text-to-speech model with 82 million parameters, released under the Apache 2.0 licence. It runs in your browser through ONNX Runtime: on your graphics card with WebGPU when that works, otherwise on the CPU with a smaller 8-bit version of the model. Long text is split into sentences and read one piece at a time in a background worker, which is why there is no need for a limit and why the page stays responsive while it works.
Why there is no character limit
Most free text-to-speech sites cap you at a few thousand characters, because every character is read by a server they pay for. Here the reading happens on your own device, so a longer text costs nothing extra. The honest limits are your device's speed and memory: the finished audio is kept in memory as you go, about 2.9 MB per minute, so an hour of audio needs around 175 MB. For a whole book, reading it a chapter at a time is kinder to a phone.
What it is not
- Not multilingual here. Only English voices are offered, because the in-browser version of Kokoro only includes English pronunciation.
- Not voice cloning. You choose from the model's built-in voices; it cannot copy someone's voice.
- Not instant on old phones. On a slow CPU the reading can fall behind playback. Leave it running and download the file when it is done.
Frequently asked questions
Is it really free and unlimited?
Yes. There is no character limit, no daily quota, no account and no watermark. The voice model runs on your own device, so a long text costs us nothing to read, which is why there is no reason to cap it. The practical limit is your device: a slow phone takes longer, and the finished audio is held in memory (about 2.9 MB per minute) until you download it.
Is my text uploaded anywhere?
No. Your text is turned into speech inside your browser tab. The only downloads are the voice model and voice files from Hugging Face the first time you use them; your text is never sent with those requests or anywhere else.
Which voices and languages are there?
28 English voices: 20 American English and 8 British English, female and male. The Kokoro model has been trained on more languages, but the browser version used here only ships the English pronunciation rules, so this page only offers English. Text in other languages will be read with English pronunciation, which is not useful.
Can I use the audio commercially, for example on YouTube?
The Kokoro-82M model and its voices are released under the Apache 2.0 licence, which allows commercial use, and the model's author says it was trained only on permissively licensed or non-copyrighted audio. The audio is generated on your device and we claim nothing over it. You are responsible for the text you have read out, and some platforms ask you to label AI-generated voices.
How long does it take?
It depends on your device. On a laptop with a working GPU (WebGPU) it is typically many times faster than real time; on a CPU it is slower, and on an older phone it can be slower than playback. Playback starts as soon as the first sentence is ready, and the page shows a time estimate for the rest.
Can I switch tabs while it works?
Yes. The reading runs in a background worker, so it keeps going when the tab is in the background. On phones, the browser may pause a tab that has been in the background for a while, or when the screen locks, so keep it in front for long texts there.
What format is the download?
MP3 or WAV, your choice. The model produces 24 kHz mono audio, and both formats keep that rate. MP3 at 64 kbps (about 0.5 MB per minute) is plenty for a single voice and is the default; 128 kbps (about 1 MB per minute) is there if you want more headroom. WAV is 16-bit and lossless but large, about 2.9 MB per minute. The MP3 is encoded in your browser with the LAME encoder when you press Download, and you can download both formats from the same reading.
Why does a word sound wrong?
Pronunciation comes from rules plus a dictionary, so unusual names, acronyms and some numbers can come out oddly. Writing a word the way it sounds usually fixes it. Punctuation matters too: commas and full stops become pauses.