Quiet Transcript records an interview in your browser, transcribes it there, plays it back beside the transcript so you can check the draft, and lets you correct it. There is no account, no upload and no server that could receive the audio even if one were asked to.
What it is for
Recordings that carry an obligation: a journalist's source, a lawyer's client, a clinician's patient, an HR investigation. For these, "the file never leaves your device" is not a feature line — it is the only lawful way to do the work.
What it is not
It is not a meeting assistant, and it does not work out who is speaking — you can type a speaker label on any line yourself, but nothing here separates voices and nothing here will. It is not a translator: it transcribes what was said in the language it was said in, and never turns it into English. It keeps nothing on its own: there is no autosave, so download the transcript before you close the tab. It is not a certified transcript either: what comes out is a draft to check against the recording.
What language it understands
Ninety-nine of them, including Russian, Ukrainian, Spanish, German, French, Portuguese, Arabic, Hindi, Mandarin, Japanese, Korean and Turkish. It works the language out from the first thirty seconds, prints what it decided beside the transcript, and lets you correct it — which transcribes the recording again with your answer forced.
English is its strongest by some distance. The small model that every visitor gets is rougher outside English and can be poor in a language it saw little of; a larger, better model is offered on the tool page and downloads an extra 80 MB only if you choose it. That is the honest shape of it: usable in a lot of languages, excellent in none, and never quiet about which language it thought it was hearing.
What is left behind
Nothing of your recording ever reaches us — there is no server here for it to arrive at — and the transcript, your corrections and the speaker labels are gone the moment you reload. Two things can stay on your own machine, and only two. The speech model itself, about 70 MB, which your browser caches so the second visit does not download it again; it is the same file every visitor gets and says nothing about you. And any recording you deliberately pressed Keep on this device for.
A kept recording is written into this browser's storage on this computer. It survives closing the tab and restarting the machine, it is never sent anywhere, and it can be played by anyone who can use that computer. The tool lists everything it is holding on the front page and deletes any of it, one at a time or all at once; clearing this site's data removes it too. Privacy says exactly what is written, where, and how to get rid of it.
How it is paid for
By one advertisement, at the foot of the tool page, from Google AdSense. It sits below everything — the machine, the transcript, the shelf, the whole explanation — never beside a control, and it is fetched only if you scroll far enough to reach it. A visit that stops before the bottom asks Google for nothing.
We would rather say that plainly than have you find it in a network panel. The trade is narrow and it is worth knowing its exact shape. Your recording, your transcript and your corrections stay on your device and are no part of it. What Google can see, when the advertisement loads, is that a browser at your address opened this page — the same thing every site with an advertisement on it can see. Privacy and Cookies say precisely what that request carries, what can be written into your browser, and what refusing all of it costs you: nothing, because the tool never waits on it.
Who runs it
Quiet Transcript is operated by AIErudit. Reach us at support@quiettranscript.com.