Troubleshooting

Most likely: Accessibility or Input Monitoring permission is missing. The hotkey needs Input Monitoring to register while another app has focus, and Accessibility to write text into that app.

Check both in System Settings, and confirm Phantom Voice is running from your Applications folder. macOS ties permissions to the app's location and treats a copy elsewhere as a different app. See Permissions Explained.

If the audio genuinely contained no sound, Phantom Voice tells you which device was silent. Check the microphone's mute switch, then the interface gain, then the device selection in Preferences.

If you are on a build older than 1.1.0 and use a multi-channel audio interface, this was a bug: a microphone on any channel other than the first produced silence. Update. See Microphones and Audio Interfaces.

Before 1.1.0 the input device you picked in Preferences was ignored, and every recording came from the macOS system default. Update from the downloads page, then re-select your device.

Older builds opened System Settings to the Microphone pane on every launch, even with permission already granted. Fixed in 1.1.0: it asks once and never opens System Settings unless you ask it to.

If it persists after updating, confirm the app lives in Applications.

That is a small model substituting a word it knows for one it does not. Add the term to your personal dictionary, and check that Preferences does not mark it not sent — the model accepts a limited number of terms.

A larger model also helps. See Choose a Speech Model.

It should not, and it is checked to prevent that. Cleanup may only remove filler, fix punctuation and capitalisation, and correct into a dictionary term. Anything else and the result is discarded in favour of your original transcript.

If you want the raw transcript regardless, turn cleanup off. See Optional Cleanup.

Transcription time scales with the model. Try Medium or Small if you dictate while your Mac is under load. See Choose a Speech Model.