Voice Setup

Voice dictation runs entirely on your Mac. No audio is sent to any server. To dictate, enable the feature and download a speech model.

Enable voice

Open Settings › Voice and toggle Voice on. Dictation stays inactive until you download a model.

Download a model

Still in Settings › Voice, pick a model and click Download. The model file is stored locally; you can delete any model you no longer use to reclaim disk space.

Model Notes
Parakeet TDT v3 Multilingual (25 European languages). Default. ~180 MB.
Parakeet TDT v2 English only. Slightly smaller than v3 at similar accuracy. ~170 MB.
Paraformer Bilingual Chinese (Mandarin + dialects) and English. ~115 MB.
Whisper Tiny 90+ languages. Broadest coverage, lower accuracy than Parakeet. ~116 MB.

Set a hotkey

The default activation hotkey is ⌘E. To change it, click the hotkey field in Settings › Voice and press your preferred combination.

Choose a mode

Two activation modes are available:

  • Toggle: press the hotkey once to start recording, press it again to stop.
  • Hold: hold the hotkey while you speak; release to stop.

Post-processing

Verne can clean up the raw transcript before inserting it:

  • Convert numbers: spoken numbers become digits (“forty-two” becomes 42).
  • Developer terms: common spoken forms are rewritten to their proper casing and punctuation (“next js” becomes Next.js, “type script” becomes TypeScript, and so on).
  • Custom replacements: add your own rules in the text field, one per line, in the form spoken form => Replacement.

Next steps