Comigo / Docs / Use Comigo

Voice

Open Settings → Voice to choose a language (English, Português (Brasil) or Español), a speech model and voice, and playback speed.

Talk with Comigo

Click the headset button to open the floating conversation, then Start talking.

• Comigo waits for you to pause, answers briefly, then listens again.
• The microphone pauses while Comigo speaks, so it doesn't hear itself. Use Interrupt to stop a reply.
• Silence sends nothing and costs nothing. Comigo never talks just to fill a pause.
• End conversation (or closing the panel) stops listening and playback.
• You can always type instead.

Your speech is recognized on your Mac: with macOS dictation when it's available, otherwise with Comigo's local Whisper model. The reply is written by your chosen AI model and normal model costs apply.

Dictate task names and goals

In New task, hover the name or goal field and click its microphone. Speech is transcribed on your Mac and added to the field; review it before you create the task. Sessions stop by themselves after one minute. No audio or dictation history is saved.

Spoken replies and suggestions

Comigo can speak its replies with a natural voice from OpenRouter (default model: Microsoft MAI-Voice). This needs an OpenRouter key or Comigo credits and costs a little per use. Test voice plays a sample.

Read new suggestions aloud is off by default. When on, new advice is shortened to two sentences, and repeats, old history, dismissed advice and paused periods stay silent.

Only the text being spoken is sent for speech. Generated audio is played from a private temporary file and deleted right after.

Local speech models

Meeting and system audio, and voice input when macOS dictation is unavailable, use Whisper running on your Mac (Apple Metal). The model is downloaded once:

After the download, transcription stays on your Mac. System audio requires Screen Recording permission.