How to use Text to Speech
- Type or paste your text, up to 20,000 characters at a time (about 3,500 words). Tap "Try a sample" to hear how it reads a time, a price and an abbreviation.
- Choose one of six voices, four American and two British, and tap Sample to hear each one before anything downloads. Then pick a speed, from Slower to Fastest.
- Tap "Make the audio". The first time, the voice downloads once (about 105 MB, or 119 MB if no other AI tool here has run in your browser) and stays there. The first sentence plays as soon as it is made.
- Listen, tap any sentence to hear it from there, then save an MP3 or a WAV. On a phone, Save opens the share sheet, so you can keep the file in Files or send it on.
Why use Text to Speech
- Natural AI voices
- Kokoro-82M, an open-source text to speech model, reads with natural rhythm and stress. Heart, Bella, Michael and Fenrir speak with American accents; Emma and George are British.
- Your text stays here
- The speech is made inside your browser. Only the voice itself downloads, once, from this site; the words you type are never sent anywhere.
- Free, with no daily limit
- Make as many recordings as you like without an account, and the audio carries no watermark. Each one takes up to 20,000 characters, so a long chapter goes in two or three parts.
- MP3 and WAV downloads
- Save a 64 kbps MP3 (about half a megabyte a minute) to share or play anywhere, or a 16-bit WAV to edit. The file is named after your first words.
- Listen while it is made
- Sentences play as soon as they are ready and the sentence being read is marked in your text. Stop keeps what is made, and Continue carries on from the next sentence.
- Other languages read aloud
- For Arabic, Urdu and other languages, the page can read your text at once with the voice built into your phone or computer, when it has one. That voice plays, but gives no file.
About this tool
Text to Speech is free, needs no account, and makes the voice on your own device, so the text you paste is never sent to a server. Its AI voices read English only. This text to speech tool, also called TTS, uses Kokoro-82M, an open-source AI voice model with 82 million parameters, published under the Apache licence. Before a voice speaks, your text is turned into sounds by a port of misaki, the pronunciation system Kokoro was trained with: dictionaries of more than 180,000 English words and names, rules for numbers, dates, money, clock times and abbreviations ("Dr.", "U.S.", "4:45 p.m.", "£52"), and a small network that guesses words the dictionaries lack. We tested the code this page runs on 118 sentences it was never tuned on: speech recognition (Whisper) wrote 110 of them back word for word, a word error rate of 1.09%, against 1.81% for the voice built into macOS on the same sentences.
Making natural speech is real computing work, and all of it happens on your device. On a fast recent computer, a one-minute recording takes a little over a minute to make; phones are slower. So the page plays each sentence as soon as it is ready and shows how much audio is made and the time left, worked out from your own device's speed; you can also stop and save what is ready. The AI voices read English only for now: no voice model we may use on this site speaks Arabic or Urdu yet, so for those languages your device's own voice reads the text aloud, without a file. To go the other way, from a recording to text, use audio to text.
Frequently asked questions
Is this text to speech free and unlimited?
Yes. This online text to speech has no sign-up, no watermark and no daily limit, because the voice runs on your own device rather than on our servers. Each run takes up to 20,000 characters (about 3,500 words); split a longer text into parts.
Is my text uploaded?
No. Your text is turned into speech inside your browser. Only the voice model downloads, once, from this site, and it is kept in your browser for next time.
Which voices are there, and can I hear them first?
Six: Heart, Bella, Michael and Fenrir (American) and Emma and George (British). Each has a Sample button that plays a short clip made with the same model, so you can choose before the download.
Can I download the audio as MP3?
Yes. Save MP3 gives a 64 kbps MP3 that every phone, player and editor opens, and WAV gives uncompressed 16-bit audio for editing. On an iPhone or Android phone, Save opens the share sheet: choose Save to Files, or send it to an app.
How long does it take?
On a fast computer, a little longer than the audio lasts. Phones are slower. The first sentence plays as soon as it is ready, and the page shows the audio made so far and the time left, measured on your device.
Can it read Arabic or Urdu?
Not with the AI voices yet: they read English. If your phone or computer has an Arabic or Urdu voice, the page offers to read the text with it at once; that voice plays but cannot be saved as a file. On an iPhone, add voices in Settings › Accessibility › Spoken Content › Voices.
Why is a word read wrongly?
Dictionary words are read as the dictionary gives them; other words, codes and web addresses go to a small network that guesses from the spelling and sometimes gets it wrong. Write the word as it sounds (for example "kilometres per hour" for km/h) and make the audio again.
Can I use it as a text reader to read aloud articles or notes?
Yes. Paste an article, a chapter or your notes and it reads them aloud, with the sentence being read marked on the page. Tap any sentence to hear it again.
Does it work offline?
Partly. The voice is kept in your browser after the first time, so later visits download nothing, and the audio is made on your device rather than on a server. The page itself still needs a connection to open.
What is the difference between TTS and an AI voice generator?
TTS, short for text to speech, is any system that reads text aloud. An AI voice generator such as this one makes the voice with a neural network trained on recordings of people speaking, so the rhythm and stress follow the sentence as a person reading it aloud would.
Text to speech kaise kare, text to voice kaise banaye? (How do I turn text into speech on my phone?)
English text paste karen, awaz chunein aur "Make the audio" dabayein, phir MP3 save kar lein. Paste the text, pick a voice and tap "Make the audio": it starts playing at once, and you can save it as an MP3 or WAV file when it is done.
Does it work on iPhone and Android?
Yes. Open the page in Safari or Chrome on an iPhone or on an Android phone or tablet, type or paste the text and tap play: the voices speak in your browser, with no app and no account. The AI voice is made on the phone, so a long text takes longer than on a computer. Save opens the share sheet, so you can choose Save to Files or send the MP3 to another app.
Built and maintained by Saad Naeem
AI engineer, Doha · Updated