Voice commands

Dictation stops at the edge of the text field. You speak an email, and then you reach for the mouse to send it; you finish a note, and reach for the keyboard to make a new one. The hands you just avoided using come straight back for the part that is not typing.

JotMic closes that gap on macOS by treating the same held key as a command channel as well as a dictation one. It is the same hold, the same release; what changes is that some of what you say is recognised as something the computer can do.

It reads the menu bar, so there is no command list

This is the part that makes it work in practice. Nearly every voice-control tool ever shipped has been a list of phrases somebody wrote by hand, one application at a time — a list that is always out of date and never covers the app you happen to be in.

JotMic does not have one. When you hold the key, it reads the menu bar of the frontmost application through the macOS accessibility API — the same menus you would open with the mouse, with their keyboard shortcuts attached. Every action that application advertises is therefore already a command. Say “new folder” in Finder and a folder appears. Say “merge layers” in a design tool and it does that instead.

Coverage is whatever the app offers, it updates when the app updates, and it needs no permission beyond the Accessibility grant that dictation already requires.

And the verbs a menu bar cannot offer

Some things are not in any application’s menus because they are not any application’s business — launching a program that is not running, or acting on the machine itself. Those are built in:

• Applications. open Slack · switch to Figma
• Browsing. new tab · close tab · scroll down
• Editing. copy · paste · undo · send it
• The machine. lock the screen · volume up · next track

Media, volume and brightness are posted as real system events rather than simulated keystrokes, so they behave exactly as the keys on your keyboard do.

Nothing fires by accident

The risk with speaking commands into a dictation tool is obvious: you say something ordinary and the computer does something alarming. Matching is therefore exact and whole-phrase against the menu titles, never a prefix and never fuzzy — the words people speak are far more often words than commands, so the bar is set where ordinary prose cannot clear it. Anything unrecognised is simply typed, which means a command that does not fire is visible as text rather than silent.

The whole layer is one switch in Settings if you would rather have dictation and nothing else.

Where this leaves your hands

The practical effect is a stretch of work with no keyboard in it: open the app, dictate the message, send it, switch to the next one, and lock the screen on the way out. For anyone managing RSI, that stretch is the point. For everybody else it is the moment dictation stops feeling like a text-entry accessory and starts feeling like a way of driving the computer.

Commands are included on every plan, including the free sixty minutes, and they work on the offline engine as well as the hosted one. See also JotMic on macOS.

Questions