Paste anything and hear it read aloud with natural voices. Adjust speed and pitch, and follow along as each word highlights.
🔒 Reading uses your device's built-in voices — the text never leaves your browser.
Listen to notes and articles on repeat, hands-free — great for retention and for resting your eyes.
Hearing your writing read back catches clumsy sentences and typos your eyes skip right over.
A quick reader for anyone who finds listening easier than reading, with adjustable speed and pitch.
This tool uses the Web Speech API's synthesis interface, which hands text to voices provided by your operating system and browser rather than to a server. That means the available voices differ between Windows, macOS, iOS, Android and Linux, and between browsers on the same machine — a voice list is a property of the device, not of the page.
Because synthesis is local, nothing you enter is transmitted. It also means quality varies enormously: some platform voices are modern neural models that sound close to natural, while others are older concatenative or formant systems that sound distinctly synthetic.
Voices load asynchronously, which is a quirk worth knowing if a voice list appears empty at first — the browser populates it after the page loads, and it may take a moment. Some platforms also require a user interaction before speech will start, part of the same autoplay restrictions that govern audio and video.
Punctuation is the main control you have over pacing. Commas produce short pauses and full stops longer ones, so text written in long unpunctuated sentences is delivered as an exhausting run-on. Breaking text into shorter sentences improves synthetic delivery far more than adjusting the rate.
Numbers, abbreviations and symbols are where synthesis most often goes wrong. Dates, currency amounts, ordinals and units are expanded according to rules that vary by voice and locale, so "1/2" may be read as a fraction or a date and "Dr" as doctor or drive. Where a reading matters, writing it out in words removes the ambiguity entirely.
Homographs are unfixable by punctuation — read, lead, live, wind and bass are pronounced by guesswork from context, and the guess is often wrong. Rate around 0.9 to 1.1 is natural for most voices; pushing much faster exposes the synthesis artefacts that slower delivery hides. Where a passage needs precise control, some platforms support SSML markup for pronunciation and emphasis, though browser support through this API is inconsistent.
Text-to-speech is genuinely useful for proofreading, since hearing text read aloud surfaces awkward phrasing, repeated words and missing connectives that the eye skips over. It also supports reading for people with dyslexia or visual impairment, and allows long documents to be consumed while doing something else.
It is not a substitute for accessible markup. A screen reader user relies on semantic structure — headings, landmarks, alternative text, form labels — to navigate a page, and reading the visible text aloud provides none of that. Adding a read-aloud button to an inaccessible page does not make it accessible, and can obscure the fact that the underlying markup needs work.
On voice cloning, the ethical and legal position has moved quickly. Synthesising a specific person's voice without consent raises personality-rights issues in many jurisdictions, several of which have introduced or are introducing specific legislation, and it is the mechanism behind a growing category of fraud. Disclosure that audio is synthetic is increasingly both an expectation and a requirement.
The list comes from your operating system. On Windows, Mac, Android and iOS you can install extra language and "natural/enhanced" voices in the accessibility or language settings — they'll then appear here.
Word timing is reported by the browser's speech engine and can lag a little on some voices, especially at high speed. The audio itself stays correct.
System voices can't be exported directly. Play the text and capture it with our voice recorder or screen recorder to keep an MP3.
Voices come from your operating system and browser rather than from this page, so the list depends on your device and platform. That is also why nothing you type is transmitted anywhere.
Homographs such as read, lead and live are guessed from context and often guessed wrongly, and abbreviations and numbers are expanded by voice-specific rules. Writing the intended form out in words is the reliable fix.