Text to Speech

Have any text read aloud by the voices on your device, with each word highlighted as it is spoken. Choose the voice, speed, pitch and volume, and see how fast the voice really talks. Which voices you get depends on your browser and operating system.

Generator Media & Files Updated Oct 16, 2026
How to Use
  1. Type or paste the text, or press a sample: a pangram, numbers and dates, a proofreading test or a longer passage.
  2. Pick a voice. The list comes from your browser and operating system; “online” voices send the text to the vendor’s service, the others run on your device.
  3. Set the speed (0.5× to 2× of the voice’s normal rate), pitch and volume.
  4. Press Speak. Each word is highlighted as the voice reaches it, when the voice reports word positions; Pause and Stop work at any time.
  5. When it finishes, Show Work gives the measured time and the voice’s real words per minute at that speed.
Input
up to 20,000 characters
× normal
1.00×
0–2, 1 = normal
1.00
100%
Samples
What Is Being Read
Words
—
Characters
—
Pieces
—
Voices
—

Worked Example

The longer passage. The sample has 6 sentences, 71 words and 397 characters. No sentence is over 200 characters, so the tool joins neighbours while they fit and sends 3 pieces of 166, 143 and 88 characters, one after another. On Windows, the device voice Microsoft David read it at 1× in 27.1 seconds: 71 ÷ 27.1 × 60 = 157 words a minute. Your voice will give its own figures, which Show Work measures.

Changing the rate. The rate is relative to the voice’s own normal speed; the specification says 2 is twice as fast and 0.5 half as fast, allows 0.1 to 10, and lets voices limit it further, which is why this tool offers 0.5× to 2×. Measured, David at 1.5× took 21.2 seconds, 201 words a minute: 27.1 ÷ 21.2 = 1.28 times faster, not 1.5. Treat the rate as a request that each voice applies in its own way. Pitch runs from 0 to 2, with 1 the voice’s normal pitch.

The common mistake: expecting a download. Speech synthesis in a browser plays straight to the speakers. The Web Speech API has no way to hand the audio to the page, so a site built on it cannot give you an MP3 of the voice. Tools that do offer one send your text to a server that makes the audio there.

Show Work

Type some text to see how it is counted and split; press Speak to measure the voice’s real speed.

Formulas

Words per minute
wpm = words ÷ seconds × 60
Measured from the first piece starting to the last one ending
Time at another rate
t ≈ t at 1× ÷ rate
A request, not exact: David at 1.5× was 1.28× faster
Pieces
piece ≤ 200 characters
Whole sentences where they fit; long ones split at commas, then spaces
Highlighted word
position = piece start + charIndex
charIndex comes from the voice’s boundary events
Words
words = runs of non-space characters
“$1,234.56” counts as one word, though a voice says several

How Browsers Speak

The Web Speech API, published as a W3C Community Group specification, gives web pages a speechSynthesis object, a list of voices and an utterance with text, voice, language, rate, pitch and volume. A browser does not have to ship voices of its own: it can pass the text to the operating system’s speech engine, and each voice reports whether it is a local service on the device or a remote one. That is why the same page offers Microsoft David and Zira on Windows, Samantha on macOS and different voices again on a phone.

While it speaks, a voice may fire boundary events with the character position it has reached, which is what lets a page highlight each word. The specification says the browser must fire them if the speech engine provides them, so voices without them simply speak without highlighting. Nothing in the API returns the audio itself; it goes to the speakers.

About This Tool

This tool reads text aloud with the voices your browser and operating system provide, highlighting each word as it is spoken. It sends the text in sentence-sized pieces so long passages are not cut off, shows which voices run on your device and which are online, and measures how fast the chosen voice really speaks at the speed you set.

It runs in your browser. The page itself never uploads your text, but voices marked “online” are cloud services of your browser’s maker, so text spoken with them goes to that service; device voices work offline. There is no download, because the browser does not give pages the audio.

It is for proofreading, for listening to long text instead of reading it, for hearing pronunciation in another language, and for anyone who finds listening easier than reading.

Related tools: Image to Text Converter (OCR), Audio Spectrum Analyzer & Spectrogram, and Tone Generator, Tuner & Hearing Test.

Frequently Asked Questions

Why are my voices different from someone else’s?

The page uses the Web Speech API, which offers whatever voices the browser and operating system provide. Windows, macOS, Android and iOS each ship their own, Edge adds many online “Natural” voices, and you can install more languages in your system settings. In Edge on Windows we saw 324 voices, of which 3 ran on the device; another browser on the same computer can show a different list.

Can I download the speech as an MP3?

No. The Web Speech API plays the voice straight to your speakers and gives web pages no access to the audio, so no in-browser tool built on it can offer a real download. To keep the audio you would have to record your computer’s sound output with a separate program.

Is my text sent anywhere?

This site never receives it. Voices that run on your device, marked without “online”, synthesise the text locally and work offline. Voices marked “online” are provided by the browser vendor’s cloud service, so the text you speak with them is sent to that service. Pick a device voice if that matters.

Why are some words not highlighted, or why is there no sound?

Highlighting depends on the voice sending word-boundary events, and not every voice does; Show Work reports how many arrived. If nothing plays, check the system volume and that a voice is listed; some browsers load the voice list a moment after the page opens, and browsers only allow speech after you press a button.

Why does it speak one sentence at a time?

Some speech engines stop long utterances partway through, so the tool sends the text in pieces of up to 200 characters, split at sentence ends, and starts each piece when the previous one finishes. It also lets the panel show which piece is being spoken.

How do I use the Text to Speech?

Simply pick your options and read the result, which refreshes the instant you change something. There is nothing to submit and nothing to wait for.

Does it cost anything or need an account?

No. The tool is completely free, there is no account to create, and it keeps working offline after the page first loads.

Is anything I type uploaded?

No. The tool works entirely on your device, so the values you enter never leave your browser.

Common Use Cases

Proofreading by ear

Hear “in in” and “their are” that the eye skips; the proofreading sample has three such mistakes in 20 words.

Listening instead of reading

At about 157 words a minute, a 1,500-word article takes Microsoft David roughly 9.5 minutes at 1×; raise the rate to get through it faster.

Pronunciation practice

Pick a Spanish, French or Japanese voice to hear a sentence in that language.

Checking how a screen reader voice sounds

Hear how a device voice reads numbers, dates and prices such as $1,234.56 and 12.5%.

Last updated: