Nothing happens when I hold the key
First check that the voice model finished downloading: the menu bar icon’s Model Status shows progress, and dictation only works once the voice model is ready. If the model is ready, check System Settings > Privacy & Security > Accessibility. Without that permission Mumble cannot listen for your shortcut at all. Toggle Mumble Dictation off and on there, then try again.I speak, but no text is inserted
If the pill appears and transcribes but nothing lands, the target app did not accept the text. Mumble then copies your words to the clipboard and shows a panel, so paste with Cmd+V. This happens when no text field has focus, when you switch apps mid-dictation, or in a few apps that handle pasting unusually. Click into the text field first and stay in the app until the text lands.The Fn key does something else
macOS has its own Fn behaviors. Open System Settings > Keyboard, and set “Press fn key to” to Do Nothing; if you use Apple’s built-in Dictation, turn its Fn shortcut off too. Or sidestep the conflict by giving Mumble a different key in Settings. How to change itA permission was denied
macOS only asks once. Open System Settings > Privacy & Security, then enable Mumble Dictation under Microphone and Accessibility. If the microphone still fails after that, unplug and reconnect your input device, and if it persists, restart your Mac.The model download failed
Click Retry; downloads resume from where they stopped, so nothing is lost. If it keeps failing, check your connection and free disk space (about 6GB in total for the models).”On-device transforms need 16GB of memory”
Macs with 8GB to 15GB run transcription on-device, but the AI layer needs either 16GB of memory or your own AI provider key. Connect one on the Model page in the sidebar to turn transforms on. See RequirementsUninstalling
Quit Mumble Dictation from the menu bar, then delete it from Applications. To remove your data too, use Delete all data in Settings first, or delete the folder~/Library/Application Support/Mumble Dictation, which holds the models, history, and dictionary.
Still stuck? Email [email protected]. We
read every one of them.