Skip to Content
Platform.Ontology

Web Speech recognition

Turns microphone audio into text.

Asks you firstCamera, mic and screenCaptureWeb browsersWhile you use it

In one line

Can turn your spoken words into text while you speak and the tab is open

Turns microphone audio into text.

Uses the microphone to transcribe speech. In Chrome the audio is sent to a Google service by default, so spoken words leave the device unless on-device recognition is chosen.

What it can and can't do

What the user does: You click the microphone or Dictate button and allow the microphone

It can

  • Turn your speech into text as you talk
  • Use the words to fill a box, search or run a voice command
  • Work on the device itself in Chrome when the language pack is downloaded

It cannot

  • Start without the microphone prompt on first use
  • Listen after you stop or leave the tab
  • Keep your audio private in Chrome by default; audio may go to a speech service unless on device mode is used
  • Work in Firefox, and only partly in Safari and iPhone browsers

How it differs between platforms: Chrome, Edge and Samsung Internet support it fully. Safari and iPhone browsers support part of it. Firefox does not.

Browser support

BrowserStatusFirst versionReleasedNote
ChromeSupported1392025-08-05webkit prefix
EdgeSupported1392025-08-07webkit prefix
FirefoxNot supportedNot availableNot listedNone
SafariPartial14.12021-04-26webkit prefix, Siri based
Safari on iPhonePartial14.52021-04-26iOS dictation, webkit prefix
Chrome on AndroidSupported1392025-08-05None
Samsung InternetSupported30.02026-05-11None

How a site asks for it

APInew webkitSpeechRecognition() or SpeechRecognition; recognition.start(); processLocally; SpeechRecognition.available(); install()
Permission namemicrophone
Permissions policymicrophone; on-device-speech-recognition
Browser compat keysapi.SpeechRecognition
How access is grantedprompt
What the user seesMicrophone prompt on first start(); Chrome shows the mic indicator; on-device mode needs a downloaded language pack.
Needs a secure context (HTTPS)Yes
Needs a user gestureYes
How long it lastsSame as microphone permission
How to take it backRevoke microphone
Inside an iframeNeeds allow="microphone"

Typical flow

  1. Create the recognizer
  2. recognition.start() from gesture
  3. Microphone prompt
  4. result events carry transcript text and confidence

Data you get

What comes back: Transcript with confidence
Shape: event
How often: continuous or one-shot

FieldTypeUnitRangePrecision
transcriptstringtextanywords
confidencefloat0 to 10 to 10.01
maxAlternativesintegercount1 to 101

Technical limits

  • Cloud processing by default in Chromium
  • Needs mic permission and gesture
  • Firefox has no support

The story

What it is

Turns what you say into text, using your microphone.

For you

You allow the microphone, then speak, and your words appear as text or trigger a command. In Chrome, your voice is usually sent to a Google service to be turned into words, so it leaves your device.

For makers

Voice typing and voice commands make forms, search and games hands free. A maker gets a speech feature without training a model. The cost is that voice audio often goes to a third party cloud, which is a privacy and legal question.

Where you see it

TypeExampleApps
accessibilityA person with a hand injury dictates a whole email with their voice and sees the words appear in the draft.Gmail, Google Docs voice typing
gamesA word game or trivia site lets you shout your answer instead of typing it, and the page checks the words.None named
educationA reading practice site listens as a child reads aloud, and highlights words they say right or wrong.Google Read Along
productivityYou say a note or a to do item into a web app, and it saves the words in a list.Google Keep, Notion web
shoppingYou say a product name into a store search bar instead of typing, and results appear.Amazon web, Google Shopping
videoA video call tool turns speech into live captions for people who are deaf or in a noisy place.Google Meet captions
smart-homeA web dashboard for your home lets you say turn on the lights and sends the command.None named

Risks

  • Spoken words may go to a cloud service
  • Always listening mode can hear private talk

What the platform says

  • Mozilla: SpeechRecognition. Developer docs describe how a page can turn spoken words into text, with limits on which browsers support it.

Good uses and abuse

Legitimate uses

  • Voice search
  • Dictation

Known abuse

  • Voice data leaving the device

Scores

ScoreValueMeaning
Sensitivity5 of 5How revealing the data is
Utility3 of 5How useful it is to honest apps
Friction2 of 5How much the user has to do
Fingerprint bitsNot ratedRough identifying entropy, when known

History

  • 2013-01, Chrome 25: webkitSpeechRecognition ships
  • 2025-09, Chrome 139: processLocally and install() for on-device models (exact version unconfirmed)

Sources

How well checked: Based on official documentation.

On this page