OBS audio setup: give it one clean input

How OBS handles audio out of the box

OBS thinks in audio sources. Open Settings, go to Audio, and you will find two global devices waiting. The first is Desktop Audio, which captures everything Windows plays through your default output: the game, the browser, your music app, the notification pings, all of it. The second is Mic/Auxiliary Audio, which captures one input device, normally your microphone. Each source gets a row in the Audio Mixer panel with a fader, a meter, and a mute button.

So if your question is simply how to add a mic to OBS: pick your microphone as the Mic/Auxiliary Audio device in Settings, confirm the meter moves when you talk, and you are capturing. That part genuinely is two clicks, and nothing in this guide changes it. What changes is what you point those two clicks at.

Where the default sources fall short

The trouble starts when your stream has more than one thing playing. Desktop Audio is a firehose: it grabs every app on the system and fuses them into a single fader before OBS ever sees them. Turn the music down and the game comes down with it. A private message ping lands on the VOD. There is no per-app split at that level, because Windows handed OBS one combined feed. Newer OBS versions can capture a single application as its own source, which helps, but every capture still dead-ends in the same place: OBS's own mixer panel.

And that panel is a broadcast tool, not a mixing console. Its filters, the gate and suppression you can attach to a source, are per-source and live inside OBS, which means they shape what the stream hears but not what you hear. To listen to the processed result you have to switch on OBS's own monitoring, which most people leave off because it arrives late and fights with everything else. In practice you end up mixing blind: adjusting faders for an audience mix you are not actually listening to, mid-stream, in a panel built for scenes rather than sound.

The patchd way: build the mix before OBS sees it

The fix is to move the mixing out of OBS entirely. A virtual audio mixer assembles the broadcast mix first, then presents the finished result to OBS as if it were a microphone. In patchd, a Windows mixer built around this workflow, every source is a strip with its own fader, meter, mute, and effects. Two strips are always there: Desktop for system audio and Communications for your chat apps. Add your microphone as a selectable strip, and pull any single app, your game or your music player, onto its own strip with a virtual cable; the walkthrough in how to route audio per app on Windows covers that move. At launch the free tier will include your mic plus one cable on top of Desktop and Communications, with three buses to route to; the Studio tier raises that to five selectable inputs and six buses in total.

Then pick a bus for the broadcast. Bus 1 works well. A bus is a virtual output that other apps see as a microphone, and each strip has a routing button per destination: light a strip's Bus 1 button and that source is on the stream, leave it unlit and it is not. Your mix is defined by inclusion, which sources you route there, so the broadcast carries exactly what you chose and nothing you did not.

Give OBS exactly one input

Now make two changes in OBS Settings, under Audio. First, set Mic/Auxiliary Audio to the patchd bus. OBS treats the bus like any microphone, except this microphone is your entire finished mix: voice, game, alerts, and stream-safe music, already balanced.

Second, set Desktop Audio to Disabled. This step is the one people skip, and it is why streams end up sounding doubled. The bus already carries your desktop sound, so if Desktop Audio stays enabled, OBS captures the same audio twice: once inside the mix and once raw from Windows. The result is a hollow, phasey echo on everything, the same doubled-audio symptom described in how to fix mic echo. One input means one copy of every sound. After these two changes, OBS's Audio Mixer shows a single meaningful row, and every level decision lives in patchd where you can see and hear it.

Keep monitoring on your headphones, not in OBS

With the broadcast handled, leave OBS's per-source monitoring switched off. Your ears get their own feed in patchd: route whatever you want to hear to your headphone hardware output, including things the stream should never get. A private call on the Communications strip, for example, goes to your headphones while its Bus 1 button stays unlit, so it reaches you and never reaches the broadcast. patchd runs over ASIO and WASAPI Exclusive, the fastest, lowest-delay audio paths Windows offers, so what you monitor feels immediate rather than delayed.

The full chain, and where patchd is today

This guide covers the OBS handoff; the complete build, including splitting every source onto its own strip and cleaning up your mic with a gate, noise suppression, EQ, and a compressor from the + Add Node menu, is in audio setup for streaming, start to finish. One nice consequence of mixing outside OBS: because the processing lives in patchd, the same cleaned-up voice reaches OBS, your recording app, and your chat app alike, instead of existing only inside OBS's filters.

To be clear about timing: the OBS-side advice here, understanding your sources and disabling Desktop Audio when another device carries your system sound, is actionable today with whatever input you feed OBS. patchd itself is in development. The free tier described above is what it will include at launch, and the only thing to do right now is join the waitlist so we can tell you the moment it is ready.