How it works
Hold. Talk. It's typed.
Three steps you can feel, and the engineering underneath them.
Hold
A global shortcut wakes the microphone. The overlay shows a live level meter so you know it is listening. Double-tap to latch and talk hands-free.
Talk
Audio goes into an on-device speech model running on your processor. Voice activity detection trims silence; your vocabulary biases spelling toward the names you taught.
It's typed
Release, and the finished text is typed at the cursor through the operating system's input layer, then your clipboard is restored. Optional transforms rewrite it on the way through.
What never happens
No audio upload. No account required. No browser extension. No analytics on what you said.