Game audio and a second voice: a practical layout for your recording.
The game and the Broker share one path into OBS: he speaks through the Windows default output device, so Desktop Audio captures him with the game, and there is no separate audio track yet. Your microphone is its own source. A track plan that keeps your voice separable, the steps in OBS, and a listening checklist.
The short answer
Three sounds, two paths. The game and the Broker share one: he speaks through your Windows default output device, so OBS Desktop Audio captures him together with the game, and there is no separate audio track yet. Your microphone has a path of its own, the Mic/Aux source. OBS can write several audio tracks into one recording, so the layout that keeps your options open is: everything on track 1 for players that play only one track, the desktop mix on a track of its own, your voice on a track of its own. Your voice can then be removed, re-levelled or replaced in an editor; his cannot be lifted out of the game’s sound. Below: the paths drawn, the track plan, the steps in OBS, and a listening checklist for the test recording.
How each sound reaches OBS
Desktop Audio captures a playback device. The Broker’s voice is in it because he speaks through your Windows default output device; the game’s sound is in it because the game plays through the same device, which is the usual case. What is on one device is one signal: one row in the mixer, one meter, one fader, one copy in the file. Adjust Broker volume in the app under Settings → Voice & capture and the game in the game; that balance is made before OBS and cannot be remade afterwards. If you switch headsets or speakers mid-session, stop and start the session again.
Mic/Aux captures a capture device, your microphone. OBS treats it as a source of its own: its own row, meter, fader, filters and track checkboxes. The app has no microphone input and no chat; any microphone in the recording is yours, through OBS, and the Broker never hears it.
The Browser source is the picture path and nothing else: the overlay shows his avatar and spoken captions on a transparent background and has no sound of its own. If you can see him and not hear him, the missing-voice guide works down the sound path from Windows to the file.
Tracks: what can be separated and what cannot
The OBS Knowledge Base describes the mechanism in two moves. In Advanced Audio Properties, opened from the gear in the mixer or from the Edit menu, each source has a checkbox per track; in Settings → Output you choose which tracks the recording contains. It adds the caveat that decides the plan: standard video players play only one audio track at a time, so everything goes to track 1 and the other tracks are for separating what an editor will want. OBS offers six.
Track
Contains
Why
1
Desktop Audio and Mic/Aux
Plays in any player; the finished mix if you never open an editor
2
Desktop Audio only
The game and the Broker, one mix: the part that cannot be split
3
Mic/Aux only
Your voice alone: remove it, re-level it, replace it
An example plan, not a rule; the track numbers are arbitrary and the file carries whichever tracks you enable.
What the plan gives you: a version without your voice (track 2), your voice by itself (track 3), and a file that still plays everywhere (track 1). What no plan gives you: a version without him, or him by himself, because his voice is on track 2 with the game. If you want a commentated recording and a clean one of the same play, that is two recordings, not one file with a track removed: record a session without starting the Broker, or press PAUSE for the stretch you want clean (PAUSE holds the Broker without ending the session; he stops looking and speaking until you press RESUME, and a paused session spends no Signal Time). The no-microphone article walks through the four combinations.
A practical layout, step by step
01Windows: both sounds on the device you listen on
Settings → System → Sound → Output: the default output device is the one you listen on, and the Volume mixer shows Signal Broker and the game unmuted on it. From the setup guide: he speaks through your Windows default output device; if you switch headsets or speakers mid-session, stop and start the session again.
02OBS: Settings → Audio
Desktop Audio set to that device by name (or Default, if the default is the device you listen on); Mic/Aux set to your microphone; Desktop Audio 2 left disabled unless it captures a second device on purpose, because two entries on one device record everything twice.
03Levels at the source, then the faders
The OBS guide’s order: set levels at the device first, then in Windows, then with the OBS fader. Here that means Broker volume in the app (Settings → Voice & capture), the game’s own volume slider, and your microphone’s gain, before the mixer faders. The Desktop Audio fader moves him and the game together.
04Advanced Audio Properties: tracks and monitoring
Tick the tracks per the plan: Desktop Audio on 1 and 2, Mic/Aux on 1 and 3. Set Audio Monitoring to Monitor Off for Desktop Audio; you already hear him, Windows is playing him to you, and monitoring into the device Desktop Audio captures is the loop that makes an echo (the doubling guide).
05Settings → Output: the tracks the file contains
With Output Mode set to Advanced, the Recording tab has an audio track checkbox for each track: enable 1, 2 and 3. Simple output mode records one track.
06Record a test and listen to the file
Start a session, play until he has spoken at least once (he watches before he speaks, and quiet stretches are normal), stop, and listen to the file in a player that lets you pick the audio track. The in-app guide asks for exactly this: check your audio meters and make a short test recording before going live. Then run the checklist below.
One wire for the two of us. Yours gets its own.
The listening checklist
Eight things to listen for in the test file, with the OBS meters open for the first. The level guidance is the OBS Knowledge Base’s, which describes the meter’s zones without publishing dB figures for them; no number on this page is a rule.
Listen for
Where
Passes when
If not
Levels
The OBS meters while the test plays
Speech sits in the upper part of the yellow zone with peaks reaching no further than the lower red; game sound sits lower in the yellow; music and alert sounds stay in the green (the OBS guide’s zones). The peak dot above the meter shows the loudest moment of the last twenty seconds, which is where clipping hides
Set levels at the source first (microphone gain, the game’s slider, Broker volume in the app), then the OBS fader
His voice not buried
A stretch where he speaks over loud game sound
His words are intelligible without turning anything up
Lower the game’s own volume or raise Broker volume in the app; the Desktop Audio fader cannot separate them, and an editor cannot either
Your voice against both
The same stretch with you talking
You and he are both intelligible; neither masks the other
Microphone gain or the Mic/Aux fader: your row is separate from his
Not doubled
Each of his lines, and one game sound you know is single
Once per line, once per sound
The doubling guide: a second desktop device, a second tap on the same device, or monitoring into the captured device
Your microphone not fed back
A stretch where you speak
No second, later copy of you
Listen to this device off in the Windows Sound control panel; monitoring off; headphones if the microphone can hear the speakers
On the track your player plays
The player’s audio-track menu
His voice is on track 1; track 2 has the mix without you; track 3 has you alone
Advanced Audio Properties track checkboxes; the Recording tab’s track checkboxes
Both channels
The two meters of a stereo source
Both light up
A source that lights only the left meter is heard only on the left: Downmix to Mono in Advanced Audio Properties, as the OBS guide suggests
Nothing else on the desktop
The whole file
No call, player or browser tab in the mix
Close them before recording: everything the device plays is Desktop Audio
Fix one thing, record again. A test recording spends Signal Time exactly like playing without a recording.
Per-application capture: the one documented alternative
OBS has a second way to capture sound on Windows: the Application Audio Capture (BETA) source, which the OBS Knowledge Base documents for Windows 10 version 2004 or later and Windows 11, from OBS Studio 28. It captures one application’s audio per source, matched by window type or by executable, and the guide says to set the global Desktop Audio to disabled in Settings → Audio when you use it, to avoid echo. In principle that is the one documented way to put the game and the Broker on different sources, and so on different tracks: one source for the game, one for the Signal Broker application.
What this site can say about it is limited. The app’s own documentation describes Desktop Audio only, and the warning in the missing-voice guide applies in full: the Broker is a different application from the game, so a scene that captures only the game’s audio has no voice; capture the Signal Broker application too, and watch its meter. Nobody in this lane has tested whether the per-application source catches his playback on your build. Treat it as an experiment with a test recording, not as a setup to go live on, and keep the Desktop Audio layout above as the one that is documented end to end.
Limits, and the next step
This page covers one Windows PC with OBS; capture cards, virtual audio cables and second PCs have their own documentation. OBS and Windows change their dialogs between versions, so a label here may read differently on yours. The plan separates your voice from the desktop mix; it does not separate him from the game, because there is no separate audio track yet. How much he says is not something a layout can set: there is no guaranteed cadence, how often he speaks depends on the game, the moment and the service, and a quiet test file may simply be a quiet two minutes. Founder Alpha is early-access software; the commentary can be wrong, repeated or badly timed, and it is sold to customers in the United States only.
Next: the pre-recording check turns this page and its neighbours into one list to run before every recording, with a plain-text copy to keep beside OBS. The OBS guide has the Browser source side, and the recorded sessions are what a Desktop Audio mix of him and a game sounds like when it works.
Limits
No level numbers are given as rules: the meter zones are the OBS Knowledge Base’s guidance, which publishes no dB figures for them, and the track plan is an example. Capture cards, virtual audio cables and second PCs are out of scope.
The overlay has no sound of its own, and there is no separate audio track yet; the Broker’s voice reaches OBS only through Desktop Audio, together with the game, and cannot be lifted out of that mix afterwards. Whether OBS’s per-application capture catches his playback has not been tested by this site.
The app has no microphone input and no chat; any microphone in the recording is yours, through OBS. There is no guaranteed cadence: how often he speaks depends on the game, the moment and the service.
Founder Alpha is early-access software; the commentary can be wrong, repeated or badly timed, and it is sold to customers in the United States only.
Founder Alpha is available to customers in the United States. Founder Alpha is early-access software. Signal Broker is an AI system. The Broker's words and voice are generated using AI. No person is speaking or writing what he says.
'/%3e%3ccircle%20cx='34'%20cy='33'%20r='5'%20fill='%23061015'%20stroke='%2377fff0'%20stroke-width='2'/%3e%3cpath%20d='M20%2045c8%204%2018%203%2026-4'%20fill='none'%20stroke='%232f8f7d'%20stroke-width='3'%20stroke-linecap='round'/%3e%3c/svg%3e){kind=small}