Text to speech

Text to speech, with MP3 download

Paste your text, pick a voice and listen to it out loud, or download it straight away as MP3. Handy for checking your own writing, for practising pronunciation, for voicing a video, and for hearing a longer article when you cannot read.

Which voices there are

Over a hundred natural neural voices in thirty languages. English alone has more than twenty voices in American, British, Australian, Canadian, Irish and Indian accents, and there are German, French, Spanish, Italian, Portuguese, Polish, Japanese, Chinese and many more. Every language has a female and a male voice, and all of them sound the same on every device, regardless of which voices your Windows or phone happens to have.

The voices are rendered on our server through Microsoft's speech service, the same voices that read aloud in the Edge browser. If you want to read offline, the end of the list has Piper voices that run inside your browser: Amy and Ryan. Their model downloads only the first time; after that reading works offline and without limits.

What you can set

The voice, the reading speed and the volume. The text is read sentence by sentence, so you can see where it is and pause or resume at any point. Save the finished recording with the Download MP3 button; once it has been rendered, downloading and replaying it cost nothing more.

Frequently asked questions

Can the speech be saved as MP3?

Yes, with the Download MP3 button, for every voice in the list. The server voices arrive as a finished MP3, the in-browser voices are converted on the spot. No recording or screen sharing needed.

How much text can it handle at once?

The free plan reads up to 5,000 characters at once, roughly three pages, and Premium up to 20,000. Split longer text into chapters; the listening is easier to control that way too. The in-browser voices have no length limit, longer text just takes longer to compute.

Is the text sent to a server?

With the server voices, yes: the text goes to our server and from there to Microsoft's speech service, which turns it into audio. Neither the text nor the recording is stored anywhere or used to train models. With the Piper voices everything stays in your browser; only the voice model is downloaded, the first time you use it.

Why are the Windows voices not in the list?

Because the browser does not let their audio out, so they could only be played, not saved. The voices in the list can do both and sound the same to everyone.

Where do the voices come from?

The server voices are Microsoft's neural voices. The in-browser voices are models from the Piper project (MIT licence).