Text to speech

Type or paste your text, choose a voice and get natural-sounding speech you can download as an MP3. The voice model runs on your own device, so your text is never uploaded and there is nothing to pay per character.

How to turn text into speech

  1. Type or paste your text into the box. Up to 20,000 characters at a time.
  2. Pick a voice. Click Hear This Voice for a short sample.
  3. Click Create Speech. The first time, your browser downloads the voice model (about 90 MB); after that it starts at once.
  4. Listen to the result and download it as MP3, WAV, M4A or OGG.

What it is good for

  • Voiceovers for videos, slideshows, tutorials and social posts.
  • Listening instead of reading: turn articles, notes, emails or study material into audio for the commute.
  • Proofreading: hearing your own writing read aloud makes mistakes and clumsy sentences stand out.
  • Accessibility and language practice: hear how English words and sentences are pronounced, at a slower speed if you like.
  • Podcast intros and phone greetings. Polish the result in the audio editor or add it to an episode in the podcast studio.

How long does it take?

Your own device does the talking, so speed depends on its processor. On a recent laptop or desktop, expect it to take about as long as the speech itself, up to two or three times that: a 150-word paragraph (about a minute of audio) takes roughly one to three minutes. Older computers and phones can be slower still, and long texts on a phone may run out of memory. The page shows the time left as it works. Keep the tab open and in front, because browsers slow down background tabs.

The voice model is a 90 MB download the first time. Your browser keeps it, so later visits start right away.

How good are the voices?

The voices come from Kokoro, an open text to speech model that sounds far more natural than the robotic voices built into most devices. Some things to know:

  • It speaks English (American and British voices). Text in other languages will be read with English pronunciation.
  • Unusual names, abbreviations and technical terms are sometimes pronounced wrong. Spelling a word the way it sounds usually fixes it.
  • Punctuation shapes the delivery. Commas and periods add pauses; a new line adds a longer one.
  • You cannot direct emotion or emphasis, and very long sentences can sound flat. Paid studio voices that run on servers are still more expressive.

Private, with no character fees

Most text to speech sites send your text to a server and charge by the character or lock downloads behind an account. Here the text is turned into speech inside your browser. Nothing you type is uploaded, there is no sign-up, and the MP3 has no watermark. You can use the audio in your own projects, including commercial ones.

About the model

Speech is made by Kokoro-82M (Apache 2.0 license) running on Transformers.js and ONNX Runtime Web. Text is converted to sounds with eSpeak NG (GPL v3).

Frequently asked questions

Is this text to speech really free?

Yes. There is no sign-up, no character fee and no watermark, because your own device does the work.

Can I download the speech as MP3?

Yes. When the speech is ready, choose MP3, WAV, M4A or OGG and click Download.

Is my text uploaded?

No. The voice model is downloaded to your browser and runs there. Your text never leaves your device.

Can I use the audio in YouTube videos or commercial work?

Yes. The Kokoro model is released under the Apache 2.0 license, which allows commercial use of what it produces. You are responsible for the text you have it read.

Which languages are supported?

English, with American and British voices. Other languages are not supported yet.

Why is the first run slow?

The first time, your browser downloads the voice model (about 90 MB). It is kept on your device, so the next time speech starts right away.