How do I record system audio on Windows without a virtual audio device?

You can record your PC’s system audio without a virtual audio device by using a tool that opens a WASAPI loopback stream — a mode built into Windows that reads the audio your computer is already playing, with nothing installed and nothing rerouted.

Most guides send you to a virtual cable or to Stereo Mix. Neither is necessary, and both have real downsides.

Why “Stereo Mix” usually isn’t the answer

Stereo Mix is the classic advice, and it fails for most people now. It isn’t a Windows feature — it’s a legacy recording input that some onboard audio drivers expose, which is why the fix is always “enable it under Recording devices, and if you don’t see it, right-click and show disabled devices.” On a lot of modern laptops it simply isn’t there.

Worse, it’s bound to one analog output path. Plug in a USB headset, connect Bluetooth earbuds, or send audio out over HDMI — the normal setup for anyone on video calls — and Stereo Mix typically stops carrying what you’re actually hearing.

The old way: virtual audio cables

The workaround people reach for next is a virtual audio device: VB-Cable, Voicemeeter, or a similar loopback driver. These work by pretending to be a speaker. You set Windows’ output to the virtual device, then point your recorder at it as an input.

That creates the same problem it does on any OS: once your output is the cable, you can’t hear the call, so you have to mirror or monitor the stream back to your real headphones. It’s a driver install, a routing job you redo whenever your default device changes, and one more thing that breaks after an update. On a managed work laptop it’s often a non-starter — installing an audio driver needs admin rights you may not have, which is the same wall people hit when their company blocks meeting bots.

The native way: WASAPI loopback

Windows has had the answer built in since Vista. An app opens an audio client on a render device — your speakers or headset — in loopback mode, and instead of pushing sound out, it reads back the exact stream being sent there. That’s WASAPI loopback, and it’s the Windows form of loopback audio capture.

What that means in practice:

• Nothing to install. The capability is part of the OS audio stack, not a driver you add.
• Nothing to reroute. Your output stays exactly as it is and you keep hearing the call normally, because the app reads a copy of the same stream.
• Any output device. USB headset, Bluetooth, HDMI, laptop speakers — it follows whatever you’re using.
• Clean digital audio. You get the stream as Windows rendered it, not a microphone’s echoey recording of your speakers.

You may already be using it without knowing: OBS Studio’s “Desktop Audio” source is WASAPI loopback, its application-audio capture uses the newer per-process loopback API, and Audacity exposes loopback inputs when you switch its audio host to Windows WASAPI. The Xbox Game Bar can capture app audio too, though it’s built for games and is picky about which windows it will record.

For meetings, a recording isn’t what you want

Capturing the audio is the easy half. If your goal is meeting notes, a WAV file of the call is a worse version of the problem — you still have to go listen to it.

That’s what Canary does with the same mechanism. It uses system audio capture to read the call audio your PC is already playing, then feeds it into streaming transcription and a live, multi-resolution rolling summary — what’s being said right now, the last few minutes, and the whole call — so you can catch up during the meeting instead of after it. No virtual device, no bot in the participant list, and no browser plugin, which also means it works the same in Zoom, Teams, Meet, or anything else playing sound.

That’s the difference from after-the-call notetakers like Otter and bot-free-but-post-meeting tools like Granola: same bot-free capture story, different moment. If you want the hands-on version of the whole approach, see how to take meeting notes without a bot.

On a Mac instead?

macOS has an equivalent. Sonoma 14.4 and later expose a Core Audio process-tap API that captures system or per-app audio with no virtual device, replacing the old BlackHole and Soundflower routing — see how to record system audio on a Mac without a virtual device. Linux does it through PulseAudio and PipeWire monitor sources.

Stay transparent

Skipping the virtual audio device is about a cleaner setup, never about capturing a call quietly. Recording and consent laws vary by region — some require only one party’s consent, others everyone’s — so tell participants when you’re capturing a meeting. “No virtual device” and “no bot” reduce the plumbing, not the courtesy.