Live transcription

MoPilot takes what the other side is saying from the operating system and your own voice from the microphone, as two separate streams. On a Mac that is a Core Audio process tap, which needs macOS 14.2 or newer because that is the release Apple added it in; on Windows it is the loopback device. Because the audio comes from the system rather than from the meeting app, the meeting app can be anything that makes sound.

Two streams rather than one is what lets the transcript know which side said what without asking the meeting platform, which it is not talking to. You can name a speaker once and it applies to everything they have already said and everything still arriving.

• Twenty languages, plus an auto-detect mode that handles more than one inside a single utterance
• Their side and yours kept apart, because they arrive as different streams
• Name a speaker once; the transcript, the note and the answers all resolve it
• Search the transcript mid-call with ⌘F
• Open it on its own with ⌥⌘\

Where it stops

• It transcribes what your computer plays and what your microphone hears — not a meeting you are not in
• Speech becomes text at a cloud provider; the audio is discarded as soon as the text returns, and no recording is written
• No speaker names come from the meeting app, because nothing is connected to it
• The Mac tap needs macOS 14.2; below that there is nothing to tap