Setup

Three questions, then it hides

The console only asks what can't be answered for you. It's not a tour, and there's no step counter.

1. Permissions

On macOS, Takumi Voice needs Microphone to record and Accessibility (or Input Monitoring) to see your key and to paste. macOS only offers these in System Settings, so the app opens the right pane for you and waits.

On Windows and Linux there's nothing to grant. Neither platform gates what a desktop app can ask for here.

2. Your key

You choose the hold-to-talk key by pressing it, not by finding it in a list. The default is a bare modifier — Right Option on macOS, Right Ctrl elsewhere — because one finger is the whole gesture and the key does nothing on its own.

Chords work too, and get swallowed, so the app you're in never sees the trigger. Nothing changes until you accept the key you pressed, so a mis-press costs nothing.

Then hold it and say something. That one gesture proves the binding and the mic at the same time.

3. Licence, if you have one

Without a licence you're on the free plan: 2,000 words a week, every feature, no account. If you've bought a licence, open Account and paste the key. See Licence and machines.

After that

Closing the window hides it instead of quitting. The product is the key; the window is scaffolding. Reopen it from the menu bar or tray, which also has a Dictate now entry for the cases where a global key can't work.