Speech to Text

Where the audio really goes

The page uses the browser's Web Speech API (SpeechRecognition). This site has no server involved: it never receives the audio or the transcript, and the text exists only in the box until you copy it. But the browser itself does not necessarily transcribe on the device. Desktop and Android Chrome stream the audio to Google's speech servers and return the words; Microsoft Edge uses Microsoft's speech service; Safari uses Apple's. That is why Chrome needs an internet connection to dictate at all.

In practical terms: your voice is handled by the browser vendor under its privacy policy, not by this website. If that matters for what you're dictating, use a system dictation feature that runs offline instead.

Browser support

Languages and the regional variant each one uses

The locale shapes vocabulary and accent expectations. Mexican Spanish is understood under es-ES and European Portuguese under pt-BR, with a few more mistakes on regional words; there's no separate entry for them. Dialects of Arabic far from Modern Standard are the hardest case.

How the live transcript is assembled

Recognition runs in continuous mode with interim results. While you speak, the engine's current guess is appended to the box in real time; when it commits a phrase, the guess is replaced by the final text and joined to what came before with a space. The engine stops itself after a few seconds of silence, and the page restarts it immediately as long as you haven't pressed Stop, so pausing to think doesn't end the session.

Punctuation and capitalisation are decided by the engine, not the page. Desktop Chrome usually inserts none, so a dictated email arrives as one long lower-case run that you punctuate afterwards; Android often adds full stops. Starting a second recording appends to the existing transcript rather than replacing it, and anything you typed by hand in between is kept.

Errors and fixes

• “Microphone access was blocked. Allow it and try again.” — click the camera/microphone icon in the address bar and allow the microphone, then press Start again.
• “Recognition error: network” — you're offline, or the browser has no speech service (see the support table).
• “Recognition error: audio-capture” — no microphone was found, or another app has exclusive use of it.
• “Recognition error: language-not-supported” — the browser's engine lacks that locale; try the US or UK English entry to confirm the microphone works.
• Nothing appears while the dot pulses: check the input device in your system sound settings, and speak closer to the microphone.
• The page must be served over HTTPS for the microphone to be available; this site is, but an embedded copy in an iframe may be blocked.