A microphone records the room. To record what the computer itself plays, a call, a stream, a video in a tab, you need its output fed back in as an input. Windows and Linux can do that already. A Mac can't, until macOS 27, without a driver.
What to pick in the share dialog, for your system.
| Your computer | Best built-in route | Nothing to install |
|---|---|---|
| Windows | Audacity, Windows WASAPI, the (loopback) device |
Chrome or Edge: Entire screen, Also share system audio |
| macOS 26 and earlier | BlackHole driver plus any recorder | Chrome 141 or newer on macOS 14.2 or newer |
| macOS 27 | Screenshot app, Include System Audio (a screen recording) | As above |
| Linux | pw-record or parecord, one command |
A Chrome tab only, not the whole system |
| ChromeOS | None | Chrome: Entire screen, Also share system audio |
| Phone or tablet | The system screen recorder | None |
On Linux: one command
Every desktop sound server keeps a monitor of each output: a copy of what is
playing, readable like a microphone. Most current distributions run PipeWire,
which records the default output with pw-record:
pw-record -P '{ stream.capture.sink=true }' playback.wav
Start it, play the sound, press Ctrl+C to stop. The file is a stereo WAV at 48 kHz.
On PulseAudio, or PipeWire with its PulseAudio layer, the same job is
parecord from the default monitor:
parecord -d @DEFAULT_MONITOR@ playback.wav
pactl list short sources lists the monitors by name (….monitor) if you
want an output that isn't the default. We ran both against a 440 Hz test tone
played into a virtual output (PipeWire 1.4.2, PulseAudio 17.0 on Debian): each
file held the tone at the level it was played, and nothing else.
On Windows: Audacity's loopback
Windows can hand any app a copy of what an output plays, through WASAPI loopback. Audacity uses it, free:
- Click Audio Setup, then Audio Settings.
- Set the host to Windows WASAPI.
- Set Recording Device to the loopback of the output you listen on, for example Speakers (loopback).
- Start the sound, then press Record.
Audacity's own tutorial has two catches worth knowing. Loopback records only while something plays, so silences are skipped rather than recorded. And Windows notification sounds on the same output end up in the take, so mute them first. Some sound cards also offer Stereo Mix as a recording device (enable it under the Recording tab of the Sound control panel), which any recorder can use. It is the same idea, done by the driver.
We have not run these on Windows; the steps are Audacity's.
On a Mac: a loopback driver, or macOS 27
Up to macOS 26, nothing Apple ships records the Mac's own output as audio.
QuickTime Player and the Screenshot app (⌘⇧5) record from an input, a
microphone. screencapture is the same: its audio options take an input device
only (we read its manual on macOS 26.5).
Apple's screen recording page lists
Include System Audio in the Screenshot app from macOS 27. That makes a
screen recording, a .mov with the picture, so for audio alone you still take
the sound out of it afterwards.
Before that, the standard answer is BlackHole, a free (GPL) virtual audio device. Sound sent to it comes out as an input any recorder can use:
brew install blackhole-2ch
Then restart the Mac. The installer needs an administrator password.
- In Audio MIDI Setup, create a Multi-Output Device with your speakers or headphones first and BlackHole 2ch second, and tick drift correction for BlackHole. Without the Multi-Output Device you record the sound but stop hearing it.
- Choose the Multi-Output Device as the Mac's sound output.
- In the recorder (QuickTime Player's New Audio Recording, Audacity, the Sound Recorder here), choose BlackHole 2ch as the input.
macOS can't change the volume of a Multi-Output Device, so set the levels in Audio MIDI Setup, and switch the output back when you are done. The setup is from BlackHole's own instructions; we did not install a driver on our test Mac.
In the browser, with nothing installed
Chrome and Edge can give a web page the sound of a tab, and on Windows, ChromeOS and recent Macs the sound of the whole computer. It goes through the screen-sharing dialog, because that is the only door a browser has to it. The Sound Recorder drops the picture at once and keeps the sound.
- Open the Sound Recorder, choose Add source, then System or tab audio.
- A panel shows the steps for your system. Click Choose what to share.
- In the browser's dialog, pick the tab that plays the sound and keep Also share tab audio on. For everything the computer plays, pick Entire screen and turn on Also share system audio. Click Share.
- The source appears as Tab audio or System audio. To record only that sound, remove the microphone with its ✕. To record a call with your own voice, keep both: each has its own level.
- Press Record, then Stop. The take is saved under Recordings. Choose Download, then WAV or MP3.
You keep hearing the tab while it records. The recording is written to your browser's storage as it goes, so a crash costs at most two seconds.
The limits are the browser's, and the recorder says so before the dialog opens:
- A whole Mac needs Chrome 141 or newer on macOS 14.2 or newer, and the browser allowed under System Settings › Privacy & Security › Screen & System Audio Recording. Older versions share a tab's sound only.
- Linux shares a tab's sound, not the whole system. Use
pw-recordabove. - Firefox and Safari share a screen but never its sound.
- Phones and tablets can't share their sound with a page.
The audio box in the share dialog is easy to miss. If the shared source shows no level, stop, share again and look for Also share tab audio or Also share system audio at the bottom of the dialog.
Which to use
- Linux:
pw-record, one line, whole system. - Windows, regularly: Audacity with WASAPI loopback. Nothing to set up beyond the host and the device.
- A Mac, regularly: BlackHole and a Multi-Output Device, once. On macOS 27, the Screenshot app if you also want the picture.
- One call or one stream, any computer with Chrome or Edge, or a laptop you can't install drivers on: the Sound Recorder. Share the tab, record, and download the WAV when you stop.
