Your First Dictation
One cycle, start to finish#
A code editor, a terminal, a commit message field, a Slack thread, a browser text box. Phantom Voice inserts into the focused text field, so whatever you would type into is where the text lands.
The indicator appears above your work and switches to its listening animation. That is your confirmation the app is actually recording, rather than the key press being swallowed by something else.
No wake word, no command grammar, no pausing between words for the recognizer. Talk the way you would to a colleague.
The indicator switches to its working animation while whisper.cpp transcribes on your Mac. Short phrases finish almost immediately. Longer passages take proportionally longer, and the model you picked changes how long, see Choose a Speech Model.
The finished text is inserted at your cursor. Not a draft in a separate panel, not a clipboard entry you paste yourself. It is in the document.
If nothing happens#
Three usual causes, in order of likelihood:
- A permission is missing. The hotkey needs Accessibility and Input Monitoring to work outside the app. See Permissions Explained.
- The wrong microphone is selected. Check the input device in Preferences, and read Microphones and Audio Interfaces if you use an audio interface.
- The app is not in Applications. macOS keeps forgetting permissions for apps run from elsewhere. See Install Phantom Voice.
More symptoms and fixes in Troubleshooting.
Make it yours#
Once the basic loop works, two settings are worth a minute each: the trigger mode if you dislike holding a key, and the personal dictionary if your work involves terms a general speech model has never met.
Next#
- The Dictation Loop — what happens in each of those states
- Dictate Into Any App — where this pays off most
- Back to the documentation home