Speech to Text
Dictate or transcribe speech live using your browser's speech recognition.
Sends a request to an online service (not AI)Not available in this browser
This browser does not support speech recognition, so this tool cannot run here. Chrome, Edge and Safari can; Firefox cannot.Your voice leaves this device
- Speech is recognised by your browser's own online service: Chrome sends audio to Google, Edge to Microsoft. This tool has no server of its own, but your audio does leave your machine.
- No transcript is stored anywhere by this site. Closing the page clears it.
- Firefox does not implement this API, so this tool cannot run there. Chrome, Edge and Safari can.
About this tool
Press start and speak: recognised text appears as you go, split into segments, and you can export it as plain text or as an approximate SRT file. Before you use it, understand what it is — this runs on your browser's own speech recognition, not on this site. Chrome and Edge stream the microphone audio to Google's or Microsoft's service, so unlike every other tool here, the audio does leave your machine. That is disclosed on the page rather than buried, because it is the one case where "free online tool" means something different from what you might assume. Firefox has not implemented the API at all, and the tool says so plainly instead of showing you a button that does nothing.
How to use it
- 1Pick the language you are speaking. Choosing the exact variant matters — Spanish from Spain and Spanish from Mexico are different models, and so are the two Chinese variants.
- 2Press start and speak normally. Recognised text becomes final a phrase at a time; the italic line at the bottom is still provisional and may change.
- 3Press stop, check the segments against what you said, then copy the transcript or download it. An approximate SRT is available if you want subtitles.
Frequently asked questions
Is my voice being sent to a server?
Yes — to your browser vendor's servers, not to this site. There is no audio server behind this tool; recognition is done by the Web Speech API, which Chrome implements by streaming audio to Google's speech service and Edge by streaming to Microsoft's. We would rather state that clearly than let you assume the recording stays local because it is a browser feature. If that is not acceptable, use a local transcription tool instead.
It stops listening after a while. Why?
Long silences make the browser end the recognition session, even with continuous mode requested. This tool detects that and quietly restarts the session as long as you have not pressed stop, so a long dictation or a meeting keeps going. What it cannot fix is the network: if the connection drops, recognition fails with a network error, and the segments already captured are kept so you do not lose them.
How accurate are the SRT timings?
Approximate, and worth being honest about. The API does not report word-level timings, so each segment is timed from when its provisional text first appeared to when it was confirmed. That makes the in and out points roughly right but often a little late at the start, and silence between phrases gets attributed to the preceding cue. For burned-in subtitles you can nudge them in the subtitle generator; for anything needing frame accuracy, align it in a video editor.