Skip to content
Local Cognitive

Voice input

Voice input transcribes speech on your computer with whisper.cpp. The audio is not sent anywhere and is deleted after recognition.

Dictate

Section titled “Dictate”
  1. Click the microphone beside Send and speak.
  2. Press the checkmark to finish. The text is added to your draft.
  3. Read it and press Send when it’s right.

Esc or the cross cancels the current dictation and keeps the rest of your draft. A recording can be up to five minutes long.

Settings

Section titled “Settings”

The small arrow beside the microphone opens voice settings (also Settings → Voice):

  • Higher accuracy: Whisper Large v3 Turbo, about 1.5 GB.
  • Faster: Whisper Small, about 465 MB.
  • Language: Auto, or a fixed language, which helps with short phrases full of technical names.
  • Choose microphone… to pick an input device.

The model downloads once, on first use, and is checked by SHA-256.

Microphone access

Section titled “Microphone access”

The first dictation asks for microphone access. If it was denied, choose Microphone permissions in voice settings, allow Local Cognitive in the system’s privacy settings, and restart the app.