Skip to content

Privacy

VibeForma uses faster-whisper and works offline: speech recognition runs on your machine.

The network is needed in two cases:

  • Once, to download the recognition model.
  • For updates.

Automatic updates are checked against a manifest signed with minisign/Ed25519. The updater keeps a backup and can roll back.

History is disabled by default. When enabled, VibeForma keeps the last 100 dictations in history.jsonl.

The dictionary can be imported and exported. Logs are rotated.