Free dictation that works offline
Windows already has a dictation shortcut, Win+H. Most people find it clunky, and the popular alternative, Wispr Flow, is a subscription. There is a third option nobody mentions: run it yourself, for free, with nothing leaving your computer.
What this actually is
A small program that sits quietly in the background. Hold one key down anywhere in Windows, in any app, and talk. Let go, and it transcribes what you said using a speech model running locally on your own machine, then pastes the text wherever your cursor was.
Why local instead of an app with a subscription
Three reasons. It costs nothing to run, forever, because there is no server on the other end billing you monthly. Nothing you say gets sent anywhere, which matters if you are ever dictating anything about a client or a number you would not want floating around online. And it works with no internet connection at all.
What it is built on
The transcription runs on faster-whisper, an open, free speech-to-text model, using a small, fast version tuned for English so it keeps up in real time on an ordinary computer. It listens only while the key is held down and processes the recording the moment you release it.
The honest limits
It transcribes after you release the key, not while you are still talking, so there is a short pause before the text appears. It does not clean up filler words on its own, and it will not learn your personal vocabulary the way a paid tool eventually might. For most everyday typing, none of that matters. For long, polished documents, you will still want to read back over what it wrote.
Setting expectations before you build or buy one
If you want this for yourself, the shape is simple: a hotkey, a local speech model, and a way to paste the result into whatever app has focus. That is the whole tool. Anything fancier than that is optional, and most of it is what turns a free weekend project into a monthly bill.
Next
Turn one idea into six posts
Another everyday tool problem, solved the same way. Once your voice is typed out, here is how to get it onto six platforms correctly.
Read the next guide