What this is
speechtotext.tools turns recorded speech into text. Drop an audio or video file or record from the microphone, choose the spoken language or let it be detected, and download the transcript as plain text, SRT or VTT subtitles, or a Word document. There is no account.
Everything runs inside your browser. The recording is decoded and transcribed on your own device, and the files you download are written straight to your disk; nothing is sent to a server. The first transcription downloads the speech engine once and the browser keeps it, so later files start right away.
Why it is built this way
Online transcription services run speech recognition on their own servers, and every minute of audio costs them something. That is why free plans are measured in minutes and accounts are required. Doing the work on your device removes that cost. It also makes the privacy claim checkable: open your browser's network panel while a file is being transcribed and you will see no upload.
Its sister site TextToSpeech.tools goes the other way, from text to speech, and works the same way: in the browser, with nothing uploaded.
What you can expect
Clear recordings of one or two speakers come out well. Overlapping voices, music under the speech, strong accents and phone-quality audio make more mistakes; the Accurate setting helps with those. Names, jargon and numbers are the most common errors, which is why every line can be played back with one click. Speakers are not labelled.
The site is free and supported by the ads on the page. The privacy policy says exactly what is measured and by whom. Every message sent through the contact page is read.
Known issues
Phones transcribe much more slowly than computers. Files longer than about an hour can run out of memory on a phone; split them and transcribe the parts. Some older browsers cannot decode every format inside video files; converting the file to MP3 or M4A first works.
The software components the tool is built on, and their licences, are listed on the licences page.