100% private on-device voice models for speech-to-text and meeting transcription on macOS. No cloud APIs, no data leaves your machine without your explicit permission.