Text to Speech

Reading your own writing aloud is the oldest proofreading trick there is, and having a machine do it is better still, because the machine reads what you actually wrote rather than what you meant. A missing word, a sentence that runs on for three lines, a paragraph where every sentence starts the same way — all of these are nearly invisible on the page and immediately obvious in the ear. The technical detail that matters most here is the chunking. Browsers implement the speech API with a practical limit on how long a single utterance can be, and past it they either cut off mid-sentence or simply stop, which is why so many text-to-speech pages mysteriously fall silent partway through a long article. Splitting the text at sentence boundaries and queuing the pieces avoids that entirely while keeping the prosody natural, since the synthesiser still sees complete sentences with their punctuation. What this cannot do is give you a file: the browser will speak the text but provides no way to capture the audio, so if you need an MP3 you need a recording tool or a cloud service, both of which mean the text going somewhere else.

Text is split into passages of about 200 characters at sentence boundaries, because most browsers silently truncate or stall on a single long utterance. The duration estimate divides the word count by 150 words per minute times your chosen speed, 150 wpm being roughly what a synthesiser produces at normal rate.

The voices available depend entirely on your operating system and browser, so the list differs between a Windows PC, a Mac and a phone, and quality varies widely between them. There is no way to download the audio as an MP3 — the browser speaks it but does not expose a recording. Firefox needs speech synthesis enabled at the operating-system level. Very long passages may still be interrupted by the browser, and pronunciation of names, acronyms and technical terms is often wrong.

A 750-word article splits into around 25 passages and takes about five minutes to read at normal speed, or three and a half at 1.4×.

1. Paste or type the text you want read.
2. Choose a voice from the list your device offers.
3. Set the speed and pitch to taste.
4. Press Speak, and pause or stop whenever you like.

• Expecting to download the speech as an audio file — the browser does not allow it.
• Assuming the same voices exist on every device; the list is your operating system’s.
• Turning the speed up so far that the point of listening is lost.
• Trusting the pronunciation of names or acronyms without checking.

Paste text and hear it read back. The voices come from your own operating system, so nothing is sent anywhere and there is no character limit, no queue and no watermark. Choose a voice, adjust the speed and pitch, and pause or stop at any point.

Writers proofreading by ear, anyone who absorbs information better by listening, and people who find long reading tiring.

• Uses your device’s built-in voices, so the text never leaves the browser.
• No character cap, no free-tier limit and no sign-up.
• Speed and pitch adjust while you listen.
• Long passages are split automatically so the speech engine does not stall.

• Proofreading — hearing a sentence exposes clumsy phrasing that reading past it hides.
• Listening to an article or a set of notes while doing something else.
• Checking how a script or speech sounds at a natural pace.
• Making written material easier to take in for anyone who prefers listening.