Audio setup for streaming, start to finish

The goal, before any software

Good stream audio comes down to two ideas. First, separation: your game, your chat app, your music, and your alerts should each live on their own channel, so you can turn one down without touching the others. Second, two mixes: one feed goes out to your broadcast, and a different feed goes to your own headphones. Those are not the same thing. You want to hear a private Discord call or a copyright-safe music source in your ears without ever sending it to your viewers.

Windows on its own will not do this for you. It tends to collapse everything into one output at one level, which is why a loud alert can bury your voice or a friend's call ends up on the recording. A virtual audio mixer is the tool that pulls those sources apart and lets you build the two mixes you actually need. The rest of this guide uses patchd, a Windows mixer built around exactly this workflow, but the order of operations is the same wherever you set it up.

Step 1: get your sources onto their own strips

In patchd, each source is a strip with its own volume slider, level meter, mute button, and effects. Two strips are always present: Desktop, which carries all of your system audio, and Communications, which carries your chat apps. Those two cover most of what a stream plays without any setup at all.

Then add the sources you want to control on their own. Add your microphone as a selectable strip. To split one specific app, your music player or your game, off from everything else, pull it onto its own strip with a virtual cable, a small piece of software that carries one app's sound into patchd. That is the move that separates game from music from browser, so a spike in one never drags the others with it. If per-app splitting is new to you, the walkthrough in how to route audio per app on Windows covers it in detail. On the free tier you get your mic plus one virtual cable (on top of Desktop and Communications); the Studio tier raises that to four cables if you want a strip for every app.

Step 2: build a bus for your broadcast

A bus is a virtual output, a clean feed you assemble for one job that other apps see as a microphone. This is where the two-mixes idea becomes real. Create a bus for your stream, then send the sources you want your viewers to hear onto it: your mic, your game, your alerts, your stream-safe music. Each strip in patchd has a row of colored buttons under a BUSheader, and clicking one lights that route in the bus's color. Bus 1 is cyan, Bus 2 is violet, Bus 3 is magenta, so a glance tells you which sources are live on the broadcast. If routing as a concept is fuzzy, see what is audio routing.

Step 3: keep your own headphone mix separate

Your headphones are a hardware output, not a bus. Route everything you personally want to hear to that output, including things the stream should never get. This is the whole point of running two mixes: your ears can have the private call and the full music library, while the broadcast bus only carries the stream-safe set. Because the broadcast is its own bus and your monitoring is a separate hardware output, turning something down in your ears does not change what your viewers hear.

Step 4: keep private audio off the stream

Handle a private group call the same way, in reverse: route it to your headphone output so you can hear it, and simply do not light its button on the broadcast bus. Nothing leaks, because a source only reaches a destination when you route it there. This is far safer than muting in a hurry mid-stream, since the private feed was never on the broadcast in the first place.

Step 5: send the broadcast bus into your streaming app

Now connect patchd to the software that actually goes live. A patchd bus can present itself to other apps as a microphone, so in your streaming or recording app you select the patchd bus as your microphone. Your broadcast software then receives one finished mix, already balanced, instead of a pile of separate device inputs you have to wrangle inside it. All of your level decisions live in one place.

Step 6: clean up the mic channel

Your voice deserves the same care as the mix. Every channel in patchd has an effects rack you fill from a + Add Node menu, and the essentials for a mic are all free. Add a noise gate first, which stays silent until you actually speak so it cuts the quiet room hiss between words, then AI noise suppression to strip out fan hum and keyboard noise, then a parametric EQ (a tone control that lets you boost or cut chosen frequencies) to shape how you sound, and a compressor to even out the loud and quiet moments so your level stays steady. They run in real time, in the order you set. For the details, see how to reduce background noise on your mic, and if you hear yourself doubled back, how to fix mic echo walks through the usual cause.

Step 7: monitor in real time

Finally, listen to the processed result as you play. patchd runs over ASIO and WASAPI Exclusive, the fastest, lowest-delay audio paths Windows offers, so the cleaned-up sound reaches your headphones fast enough to feel immediate rather than delayed. Talk, watch your mic meter move, and nudge the volume slider until it sits right against the game and music.

Optional: change your voice

If you want a character voice on top of a clean signal, patchd's Persona voice changer applies a voice to your mic as the final stage after the effects rack, and switching is hotkey-first so you can flip voices live. The dedicated voice changer for streaming setup guide covers where it sits in the chain and how the voices behave.

Where patchd is today

patchd is built for exactly this: separate strips, color-coded routing, a real effects rack on every channel, and a free tier that is the full mixer rather than a stripped-back demo. It is in development now. Join the waitlist and we will tell you the moment it is ready to install.