How to separate game audio and voice

First, confirm this is the problem you have

Open your last recording in an editor and try to turn the game down without touching your voice. If you cannot, because both live inside one waveform, you have the baked-track problem. It shows up in a few familiar ways: a VOD where the game drowns out a great callout, a clip where you laughed over the exact sound effect that made the moment, or an edit session where fixing your mic level also drags the game level around with it.

None of that is a settings slider you missed in the editor. Once game and voice are summed into one file, no software can cleanly pull them apart again. The fix happens before you hit record, not after.

Why game and voice end up baked together

The usual cause is recording the blended output. Most capture setups default to grabbing what your headphones hear, and what your headphones hear is a mix: game, voice chat, browser tabs, notification pings, all summed into one stereo signal. Record that signal and you have recorded a finished decision. The balance between game and voice at the moment of capture is the balance forever.

A single mic track mixed in live has the same issue from the other side: if your recorder adds your mic into the desktop capture instead of writing it to its own track, one loud keyboard smash or one mistimed shout is welded into the take.

The fix, in one sentence

Route game audio and your microphone as separate sources to separate destinations, then record each destination on its own track. That is the whole idea, and everything below is just two ways of wiring it up. If routing is a fuzzy concept, the primer on what audio routing is makes the rest of this page click.

The Windows-side way, and where it runs out

You can get partway there with what Windows ships. It is fiddly but honest work, and for a simple two-track recording it can be enough:

1. Give the game its own output device. Open Settings, then System, then Sound, then Volume mixer, and set the game's output to a device that is not your main headphones. In practice this means a virtual audio cable, a software device that exists purely to carry sound from one app to another, because you need something your recorder can capture, not a second pair of speakers.
2. Restart the game. Plenty of games only read their output device on launch and ignore the change until you relaunch.
3. Add two sources in your recorder. Add the cable as one audio source and your microphone as another, instead of a single desktop capture.
4. Put each source on its own track. In OBS that is Advanced Audio Properties: tick track 1 for the game source, track 2 for the mic, and record to a multi-track format like MKV. Your editor will then see two independent waveforms.
5. Get the game back in your ears. Routing the game to a cable means you no longer hear it, so you have to monitor the cable source through your recorder or wire a listen path yourself.

The limits show up fast. The volume mixer moves whole apps only, some apps refuse to follow it, there are no meters to check what is actually flowing, and the monitoring step is a workaround stacked on a workaround. The full tour of those limits is in how to route audio per app on Windows.

How patchd splits them

patchd is a virtual audio mixer, a piece of software that sends sound between your apps and devices, and this exact split is what its layout is built around. Sources are strips on the left. Destinations sit on the right: your hardware outputs and your buses, where a bus is a virtual output other apps see as a microphone. Each destination owns one fixed color, and you route a strip somewhere by clicking that destination's colored pill on the strip.

1. Pull the game onto its own strip. Set the game's output to the patchd virtual cable in the Windows volume mixer. The cable shows up in patchd as its own strip with a live meter, so you can see the game flowing the moment it makes sound.
2. Your mic is already its own strip. Select it as an input and it gets a fader, a meter, and its own routing pills, completely independent of the game.
3. Route each strip to its own bus. On the game strip, light the Bus 1 pill, cyan. On the mic strip, light the Bus 2 pill, violet. Two sources, two destinations, zero overlap.
4. Route both strips to your headphones too. Pills are per destination, so a strip can feed your ears and a bus at the same time. You hear the full blend live while the recording stays split.
5. Record the buses. In your recorder, add Bus 1 as one audio source and Bus 2 as another, one track each. Your editor gets a clean game track and a clean voice track.

The monitoring problem from the Windows path simply does not exist here, because hearing the game and recording the game are two different pills on the same strip.

Keep private audio off both tracks

A split recording also solves the quieter disaster: Discord calls, notification sounds, and whatever your browser decides to autoplay. In patchd, voice chat lives on the fixed Communications strip and everything unrouted lives on the fixed Desktop strip. Leave their bus pills unlit and none of it reaches Bus 1 or Bus 2, so it never touches the recording, even though you still hear it in your headphones. The same unlit-pill trick is the whole fix in stream hearing your Discord audio, and if you are building a full capture setup, the streaming audio setup guide walks the complete picture.

The short version

• Never record the blend. One summed track is one unfixable balance.
• Separate sources, separate destinations. Game to one device or bus, mic to another.
• One track per destination in your recorder, so the editor sees independent waveforms.
• Keep chat and notifications off both recording destinations entirely.

patchd makes the split visible instead of theoretical: the game on its own strip, your mic on another, each lit to its own colored bus, and private audio staying dark. It is in development now. Join the waitlist and we will tell you the moment it is ready to install.