Your voice. Everywhere you type.

You need macOS 14 Sonoma or later on an Apple silicon Mac — an M1 or anything newer. Dictation is English-only for now. Intel Macs, Windows, iPhone and Android are not supported. The first launch downloads a speech model of about 450 MB.

Sign in with Google and the download starts. VaniQ works fully for seven days — every app, no limits — and you never enter a card to start. When the week ends, dictation pauses until you choose a plan: $9.99 a month or $99 a year. Decide early and you lose nothing: subscribe during the week and the first charge lands when your free week would have ended anyway.

Differences you will feel on the first day: you hold one key instead of turning a mode on and off; the text lands in the app you are already in rather than in a window you then copy from; the ums and uhs are taken out; and the names you add to Your words come out spelled your way. If you write code, VaniQ also spells the names in your project correctly. The honest test is your own week of writing, which is what the trial is for.

Dictation does, once the speech model has downloaded. Your account and the periodic subscription check need a connection; a paid subscription keeps working for 14 days offline after one falls due, so a flight or a bad hotel connection is not a problem.

Nearly. VaniQ writes into the focused text field in apps like Mail, Messages, Notes, Slack, browsers, editors and terminals. Password fields and a small number of apps that deliberately block automated input will not accept it. Try your own apps during the trial — and note that VaniQ never presses Return, so nothing sends itself.

Yes, and it treats them kindly. A Bluetooth headset microphone switches AirPods into phone-call mode, which makes music sound flat and gives the speech model a thin signal. So when your AirPods are connected and your MacBook is open, VaniQ listens through the Mac’s own microphone instead. If you would rather use the AirPods’ microphone, or any other, choose it from the Microphone menu.

Accuracy depends on your voice, your microphone, the noise around you and the words you use, so read anything important before you send it. On our own 20-recording test set an M1 scored 0.00% word error on ordinary prose, and 1.77% overall once project vocabulary was matching — 5.96% without it, because most of those errors are technical names. That corpus is clean studio audio, so it flatters the number. It is a measurement of our test set, not a promise about your voice.

Common ones, usually. Unusual ones — a colleague, a client, a company, a medicine — the speech model can only spell by ear. Add them to Your words, in the menu bar, and VaniQ writes your spelling whenever it hears something close. On our test sentences that restored five of eight names the model had misspelled and changed nothing else. A name heard too far from its spelling can still slip through, so read important names back before you send.

It is transcribed on your Mac and put where your cursor is. The audio is not uploaded, the transcript is not uploaded, and neither is used to train anything. Your usage totals are kept on your machine.

Both from your account page. Your subscription is not tied to one machine — sign in on any Mac you own and it works, and the account page lists the ones you have used so you can tidy it up. Cancelling stops future renewals; you keep access until the end of the period you already paid for.