Text to Speech
Uses the voices already installed in your browser and operating system — free, offline for most voices, and nothing is sent anywhere.
0 words · about 0s
Rate 1.0x
Pitch 1.0
Volume 100%
Read any text aloud using the voices already installed on your device. Useful for proofreading, for accessibility checks, or for hearing how a script sounds before recording it.
This uses the speech synthesis engine already built into your operating system, which is why the available voices differ between Windows, macOS, Android and iOS, and why the same text can sound quite different on two machines. Nothing is generated on a server and nothing is uploaded.
Rate and pitch are worth adjusting for the job: proofreading your own writing by ear catches sentences that do not parse, and a rate slightly below normal is easier to follow than the default. Because the audio is produced by the system rather than by a file, there is no download — the synthesised voices are licensed to your device, not to this page.
The voice list is the part that surprises people, because it is not a property of this page at all. It is whatever your operating system has installed, which is why the same text sounds different on Windows and macOS, why a phone often has better voices than a laptop, and why a language you expect may be missing entirely. Adding voices is done in the system settings rather than here — Windows under Speech, macOS under Spoken Content, Android and iOS under Accessibility — and a voice installed there appears in the list on the next page load.
Reading your own writing aloud is the use that justifies the tool. A sentence that has lost its verb, a clause that never closes, a paragraph that repeats a word four times: all of them are easy to read past on screen and impossible to miss when heard, because the eye silently repairs text that the ear does not. A rate slightly below the default makes this easier rather than harder to follow, and the highlighting keeps your place so a problem can be found rather than merely noticed.
There is no download, and the reason is licensing rather than a missing feature. The synthesised voices belong to your device — they are licensed to the operating system, not to this page — so the API plays audio but does not hand back a file to save. Anything that produces a downloadable narration is generating the audio on a server under its own licence, which is a different arrangement with different terms. If a recording is what you need, capturing the system audio locally is the honest route, and the licence attached to the voice still applies to whatever comes out.
The Web Speech API takes plain text and nothing else — there is no SSML support, so the pauses, emphasis and phonetic spellings that a server-side synthesis service accepts have no equivalent here. What punctuation you write is the only prosodic control available: a comma produces a short pause and a full stop a longer one, which is enough to break up a list that would otherwise be read as a single breathless run. For an unusual name, spelling it the way it sounds is the only way to steer the pronunciation, and the trick is worth remembering when proofreading a document full of proper nouns.
Some browsers load voices asynchronously — wait a moment and reopen the list. On Linux you may need a speech synthesis engine installed at system level.
No. The Web Speech API plays audio but does not expose a recordable stream, so there is no file to save.
The page does not transmit it. Note that some operating systems use a cloud voice for synthesis, in which case the OS itself may send text to its own service.