Before you start: this is not private.
Live voice typing uses your browser's built in speech recognition. In Chrome and most Chromium browsers, that sends your microphone audio to Google's servers to convert it to text. Your words leave your device.
FastScribe's own transcription does not work this way: an uploaded recording is processed without being sent to a third party for speech recognition. If you want the private route, record your audio and upload the file instead of using this page.
ready · 0 words
Dictation always needs a cleanup pass. Read it back and fix it before you rely on it.
Your microphone is released the moment you press stop or leave this page.
How to use it
- Read the privacy note. This tool works differently from the rest of the site.
- Press start and allow the microphone. Words appear as you speak, with the unsettled ones shown differently.
- Stop when you are done, edit the text, then copy it or download it as a .txt file.
For anything you need to keep, record it and upload the file. It comes back more accurate, with timestamps, and without being handed to anyone else for recognition.
Transcribe a recording instead →Frequently asked questions
Is this private?
No, and this is the one page on the site where that is true. Your browser does the recognition, and in Chrome and most Chromium browsers that means your microphone audio is sent to Google's servers to be turned into text. If the recording matters, use FastScribe instead: upload the file and it is transcribed without being handed to a third party for recognition.
Why does it not work in Firefox?
Firefox has never shipped the speech recognition part of the Web Speech API, so there is nothing on the page for it to use. Chrome, Edge and Safari all work. If you are on Firefox, record your audio and upload it instead.
Why does it stop on its own?
Browser speech recognition ends a session after a pause, which is how the API is built. The page restarts it for you while dictation is switched on, so you can keep talking. If it fails to restart several times over it stops and tells you rather than looping in the background.
How accurate is it?
Good enough for a note, and worse than transcribing a recording. Live recognition works with only the words so far, while transcribing a finished file lets the model use what comes afterwards to settle what it heard earlier. Expect to fix punctuation and names either way.
Can I dictate a long document?
You can, but it is the wrong tool for it. Long dictation is where the accuracy gap hurts most and where sending everything to a third party matters most. Record yourself instead and upload the file.