Text to speech

How it works

Type or paste your text, choose a voice and press Generate. The text is split into short parts of a sentence or two, and the first part starts playing as soon as it is ready while the rest is generated behind it — you do not wait for the whole script before hearing anything. When it is done, download the audio as MP3 or WAV, or download subtitles as an SRT file timed to the audio.

Everything happens on your own device. The first time you use it, the voice engine is downloaded once and kept by your browser; after that it starts almost instantly and keeps working without an internet connection. Your text is never sent anywhere, which is also why there is no character limit, no queue and no account: there is no server doing the work that would need any of those.

Voices

There are sixteen English voices: nine American female, three American male, two British female and two British male. Heart is the default and the most natural for general narration; Bella is brighter, Nicole is soft and close, Emma and George give a British read. Every voice has a short listen button in the voice menu, which works before anything is downloaded, so you can choose by ear.

Speed goes from 0.7× to 1.5×. Slower is useful for learners and for dense material; 1.1× to 1.25× suits short social videos and anything people will listen to on the move. The voice and speed you pick are remembered on this device.

Pauses, paragraphs and pronunciation

A blank line between paragraphs gives a short natural pause, and a line on its own — a heading or a list item — gets a brief pause too. For an exact pause, type [pause 1s] or [pause 2.5s] anywhere, or put the cursor where you want it and use the Pause button. Pauses are the easiest way to leave room for music, a cut or a slide change.

Numbers, dates, times, money, percentages, common units, phone numbers and titles such as Dr. and Mr. are read in their spoken form. Unusual names and brand names are sometimes mispronounced; the quickest fix is to spell them the way they sound — "Nguyen" as "Win", "Siobhan" as "Shivawn" — and generate again.

Fix one sentence without redoing everything

After generating, the Parts list shows each piece of the script with its length. Click a part to play from there. If a sentence needs rewording, edit it right in the list and press Save & regenerate: only that part is generated again, and every unchanged part is reused instantly. The same happens if you edit the main text and press Generate — only what changed is redone.

This makes long scripts practical. Revise the third paragraph of a ten-minute narration and you wait seconds, not minutes.

Files in, audio and subtitles out

Import a TXT, Markdown, Word (DOCX) or PDF file with the Import text button, or drag it onto the text box. The text is extracted on your device and dropped into the editor, where you can trim headers, page numbers and footnotes before listening. PDFs that are only scanned images have no text layer to read.

MP3 is the right download for sharing and listening. WAV is lossless and is the better choice if the voice goes into a video or audio editor for further work. The SRT download gives you subtitles that line up with the audio, ready for a video editor or a subtitle tool.

Speed on different devices

On most recent computers the audio is generated several times faster than it plays, so playback starts within a second or two and never has to wait. Older computers and most phones generate more slowly than real time; the tool then waits until enough audio is ready to play without stopping and tells you how long that will take. Downloads work the same everywhere — they just take longer on slower devices.

The very first generation on a computer can take up to a minute of one-time preparation after the download. It happens once; the next visit starts straight away.

Frequently asked questions