System audio capture is recording whatever sound a computer is already producing through its own speakers or headphones — the other side of a video call, a video, a podcast — as its own audio source, separate from whatever a microphone happens to pick up in the room. A meeting recorded with both running at once ends up as two tracks: the room, from the microphone, and the call, from system audio.
Why a meeting needs both, not just one
A laptop's microphone hears the person sitting in front of it clearly and everyone else on the call faintly, filtered through the same speaker that's also playing them back. Recording system audio alongside the microphone instead of relying on the mic for everything means the other side of the call is captured directly from the source generating it, not picked up secondhand through a speaker and re-recorded by a microphone a few feet away. Coii AudioNotes records the two as separate tracks for exactly this reason: the room stays distinct from the call instead of the two blending into one another before either is transcribed.
The permission it asks for
macOS treats capturing what's already playing as the same category of access as recording the screen, so an app doing it asks for that permission the first time — even though nothing about the recording involves video. It's a one-time prompt, not a recurring one, but it's worth knowing what the request is actually for: the audio, not anything visible on the display. Declining it doesn't stop a recording outright; it just means only the microphone track gets made, with the other side of any call missing from it.
What it sounds like when it's missing
A recording made from the microphone alone still captures the other side of a call, just secondhand — through a speaker, back through a mic, with whatever room noise sits in between. It's audible in the transcript before it's audible in the audio: a speaker who comes through faint, words the transcription is less confident about, occasional gaps where a quieter voice didn't carry across the room clearly enough to be picked up at all. System audio capture removes that middle step entirely by recording the call's audio at the source instead of after it's already been played and re-recorded.
How it differs from a bot's feed
A meeting bot gets its audio by joining the call itself and asking the platform for a feed, which only works for calls happening inside an app the bot knows how to join. System audio capture doesn't care what's making the sound — a video call, a video file, music playing in another window — because it captures at the level of the computer's own audio output rather than inside any one application. That's also its limit: it can only ever capture sound the computer itself is playing, so a call taken on a phone sitting next to the laptop, or audio routed to a second device entirely, never reaches it.
Where it fits into a recording
On its own, system audio capture is just an audio source, the same as a microphone is — it doesn't transcribe or summarise anything by itself. What turns it into a finished meeting record is what happens after: a notetaker transcribing both tracks and telling speakers apart within them, which Coii AudioNotes does without either track leaving the Mac, and what a private, offline recording setup looks like more generally covers the rest of that setup. It's one piece of what an AI notetaker actually does with a recording, not the whole of it.