System audio capture, also called loopback or desktop audio recording, takes the sound your computer is already sending to its speakers or headphones and records it as a signal in its own right. That stream carries the other people in a video call, a playing video, or another application producing sound on the machine.
A microphone records the air in the room, so it picks up your own voice and whatever else is nearby. System audio is the output stream the operating system routes to your speakers, so it carries the voices arriving over the network. That is why a recording made with the microphone alone, while everyone else is on headphones, usually contains only your side of the conversation, and why a desktop recorder generally needs both sources: the microphone for the person at the keyboard and system audio for the remote participants.
The recorder asks the operating system for a copy of the output stream, then stores it alongside or separately from the microphone track. Separate tracks help with speaker labelling and later editing, while a single mixed file is simpler to move around. macOS and Windows expose capture through different system interfaces, and older workarounds relied on a virtual audio driver you installed yourself, so instructions written for an earlier version may no longer match what you see.
Confirm whether a tool captures system audio at all, whether it records one application or the whole machine, whether tracks stay separate, which file formats it writes, whether a transcript is produced, and where processing happens. Support also depends on your operating system version, so read the current documentation for the tools you compare rather than a general claim.
Permission handling differs by platform. On current macOS versions a recorder asks for permission to capture audio from other applications, and microphone access is a separate prompt. On Windows, the privacy settings gate microphone access, while capture of the output stream is handled by the audio system and generally does not raise a separate prompt. Recording other people also raises consent questions that vary by jurisdiction and by workplace policy, so tell participants you are recording and check the rules that apply to you.
Cue is a desktop voice AI for macOS 13+ and Windows 10+. Its meeting capture is a recorder you start on your own machine, so nothing joins the participant list and the operating system prompts still govern what can be recorded. Cue can recognize meeting context and offer the recorder, but recording begins only after you choose Start; a progressive transcript appears during longer meetings, and structured notes with action items are available after you stop. Cue is free to start; Cue Plus is $19.99/month.