Hold-to-talk
The default dictation hotkey is Right Option (⌥). Hold it while speaking, then release. AFK captures the microphone input, runs the speech model, applies your text settings, and sends the result to the focused app.
The hotkey is system-wide, so there is no AFK window to find first. Accessibility permission is required for the global event tap and for cross-app text insertion.
Hands-free behavior
You can lock a recording without keeping the hold key down. While holding the dictation key, press the configured start sequence; the default sequence is A + F + K. Speak for as long as you need, then press the configured stop sequence with the dictation key; the default is Right Option (⌥) + Return.
Esc cancels an active recording. The hold key, start sequence, and stop sequence can all be changed in Settings. Avoid combinations already used by macOS or another app.
Where the text goes
| Focused target | AFK behavior |
|---|---|
| Editable field | Pastes the transcript into the field. AFK briefly uses the clipboard for the paste, then restores the previous clipboard contents. |
| Unknown target | Attempts the paste but keeps the transcript on the clipboard as a safety net. |
| Clearly non-editable target | Copies the transcript to the clipboard so you can paste it where you want. |
| Always copy enabled | Skips insertion and puts every finished dictation on the clipboard. |
If insertion fails, AFK also falls back to the clipboard. A dictation is recorded in local History after the result is accepted.
Cleanup and dictionary
In Instructions or Settings, you can turn on sentence capitalization and removal of filler words such as “um” and “uh”. These are post-processing options applied before the result is inserted and stored.
The Dictionary is a list of custom say this → write that replacements. Use it for names, product terms, or spellings the model should consistently write your way.
Language and model choices
The default model is Parakeet v3 Multilingual. The app's language setting steers the decoder toward the selected language's alphabet and applies to the next utterance. You can instead select the English-only Parakeet v2 model; model changes apply after the next app launch.
The supported v3 language choices are English, Bulgarian, Croatian, Czech, Danish, Dutch, Estonian, Finnish, French, German, Greek, Hungarian, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Spanish, Swedish, and Ukrainian.
Small practical limits
- Recordings shorter than 0.3 seconds are ignored.
- An optional maximum recording length can auto-stop a recording after 30 seconds, 1, 2, 5, or 10 minutes. The default is off.
- Microphone permission must be authorized before AFK can record.
- The speech model must be ready before a dictation can be transcribed.
Need to set it up? Follow the first-run guide. If the hotkey is silent, start with permissions troubleshooting.