Speech-to-Text
Speech recognition.
Loading tool…
How dictation works
Speech recognition turns an audio signal into words. The audio is sliced into short frames, converted into a spectral representation, and matched against an acoustic model that maps sound patterns to phonemes, while a language model weighs which word sequences are plausible. That second part is why recognisers correct themselves mid-sentence: a later word can make an earlier guess unlikely.
Accuracy depends on things you control more than you might expect. A close, consistent microphone position matters more than an expensive microphone. Background noise, crosstalk, and reverberant rooms hurt badly. Accents and technical vocabulary are handled unevenly, and punctuation usually has to be dictated explicitly. Expect to edit the transcript — recognition gets you most of the way, not all of it.
Why we built this tool
Dictation is faster than typing for first drafts, invaluable for anyone with a repetitive strain injury, and the quickest way to capture a thought while your hands are busy.
It is worth knowing where your voice goes. Browsers implement speech recognition through the Web Speech API, and in some browsers — Chrome in particular — that API sends audio to the vendor's servers for processing, while others recognise on-device. We add nothing to that: this tool has no server of its own, so no audio and no transcript ever reaches us, and the resulting text stays in your browser until you copy it out. Check your browser's implementation if the material is highly sensitive.
Tips and common mistakes
- A close, consistent microphone position matters more than an expensive microphone.
- Dictate punctuation explicitly, and speak in complete sentences rather than fragments.
- Background noise and reverberant rooms hurt accuracy more than accent does.
- Always edit the transcript; recognition gets you a draft, not a finished text.
Quick start: using Speech-to-Text
- Grant microphone permission when the browser asks.
- Position the microphone close and consistent, in as quiet a room as you can manage.
- Start dictating in complete sentences, saying punctuation aloud where needed.
- Watch the transcript build and pause if it drifts badly.
- Copy the text out and edit it — no recogniser produces a finished draft.
Clean the transcript with Remove Extra Spaces and Basic Grammar Checker, or record the audio itself with Audio Recorder.