Speech to Text at a glance
- What it does
- Dictate into your browser and get text back in real time using built-in speech recognition.
- Where it runs
- Entirely in your browser — no data is uploaded
- Works offline
- Yes, once the page has loaded
- Cost
- Free, with no account and no usage limit
- Category
- Text Utilities
How to use dictation
- Allow microphone access when the browser asks. Nothing works without it.
- Press start and speak at a natural pace. Text appears as you go.
- Say the punctuation - "comma", "full stop", "new paragraph" - since recognition rarely infers it.
- Stop, then edit. Expect to correct names and technical terms.
- Copy the transcript when you are done.
Browser support is uneven. Chrome and Edge implement this well; Safari's support is partial; Firefox does not enable it by default. If the button does nothing, your browser is the reason.
Getting better accuracy
Modern recognition reaches the mid-to-high nineties on clear speech in a quiet room, and falls sharply outside those conditions. What actually helps, roughly in order of impact:
- A better microphone. A headset close to your mouth beats a laptop microphone across the room by a wide margin - a bigger improvement than anything else on this list.
- A quiet room. Background conversation is far worse than steady noise, because the system tries to transcribe it.
- Normal pace and full sentences. Over-enunciating hurts. Context is what disambiguates homophones, so complete sentences beat isolated words.
- The right language setting. Selecting the correct regional variant materially improves accuracy for accented speech.
Names, jargon and acronyms remain the weak point. Dictate the prose, then fix the proper nouns.
Punctuation and formatting commands
| Say | Get |
|---|---|
| "comma" | , |
| "full stop" or "period" | . |
| "question mark" | ? |
| "exclamation mark" | ! |
| "new line" | A line break |
| "new paragraph" | A blank line |
| "open quote" / "close quote" | " " |
Support varies by browser and language. Where a command is not recognised, the word itself appears in the transcript - which is at least easy to find and remove afterwards.
Where your voice actually goes
Worth being precise about, because it is the one place where this tool differs from most of the site.
The Web Speech API does not specify where recognition happens. In Chrome and Edge, audio is streamed to the vendor's speech servers and text comes back - so Google or Microsoft processes your speech under their own privacy policies. Some platforms perform recognition on-device, particularly on newer mobile hardware and on macOS with dictation downloaded locally.
What is certain is that we receive nothing: no audio, no transcript, no recording. The transcript exists only in your browser tab until you copy it.
The practical advice: do not dictate confidential material through browser speech recognition unless you know your platform does it on-device. For sensitive dictation, use your operating system's offline dictation mode.
Where dictation beats typing
- First drafts. Most people speak at 120-150 words per minute and type at 40-60. Dictating a rough draft and editing it is often faster overall.
- Notes after a call, while the detail is fresh and typing would slow you down.
- Accessibility. Repetitive strain injury, limited hand mobility, or any situation where typing is painful.
- Hands-free capture - cooking, driving as a passenger, working with your hands.
- Transcribing your own recording by playing it back near the microphone. Crude, but it works.
Frequently asked questions
Your browser probably does not support the Web Speech API. Chrome and Edge work; Firefox does not enable it by default and Safari's support is partial. You also need to have granted microphone permission, and the page must be served over HTTPS.
Not by us - we receive neither audio nor transcript. In most browsers the audio is sent to the browser vendor for recognition and handled under their privacy policy. Nothing is stored on our side because nothing reaches it.
Usually not, since recognition typically happens on the vendor's servers. Some newer devices support on-device recognition, which does work offline.
Whatever your browser's recognition service supports - typically several dozen languages with multiple regional variants. Choosing the variant closest to your accent noticeably improves accuracy.
Say it aloud: "comma", "full stop", "new paragraph". Recognition does not infer punctuation from your intonation.
Nothing you enter here leaves your browser
Speech to Text does its work in JavaScript running on your own device. The page loads once, and after that there is no upload step and no server involved — which matters here because unpublished drafts, contracts and internal documents stay on your machine.
You can verify this rather than taking our word for it: load the page, disconnect from the internet, and the tool keeps working. Our privacy policy sets out what is and is not collected, and this guide explains why the distinction matters.