Splitting Game, Mic and Discord Audio Into Separate OBS Tracks

Last spring I recorded a two hour co-op session for a friend's channel, and in the edit I found a Discord notification ping sitting on top of the one funny line of the night. The ping, the game soundtrack, my mic and my friend's voice were all baked into a single stereo track. I could not lower one without lowering all of them, so the line went in the bin.

That is the most common audio complaint I hear from people who record games: the recording sounds fine live, then turns out to be uneditable. The fix is not a new microphone. It is spending twenty minutes in OBS Studio so that each sound lands on its own track, which gives you a mixing desk after the fact instead of a flattened recording you have to live with.

Why one track ruins a good recording

A stream needs a single finished mix because viewers only hear one thing. A recording does not. When you send the same mix to disk that you send to Twitch, you are making every mixing decision in real time, while also playing the game, which is the worst moment to make them.

The trouble shows up in three predictable places. Game music turns out to be a licensed track that gets the upload muted or claimed. A teammate's voice chat is twice as loud as you, or contains something they would rather not have published. Your own voice sits under an explosion and cannot be lifted without the explosion coming up too. Each of those is a two minute job with separate tracks and an impossible one without them.

OBS has supported up to six audio tracks in a recording for years. Most people never touch it because the setting hides behind the Advanced output mode, and the default Simple mode writes everything to track 1.

The layout I use, and why track 1 stays a full mix

The advice you see most often is to put every source on its own track and nothing else. I disagree with the "nothing else" part. Keep track 1 as the complete mix, exactly what the audience would hear, and use tracks 2 to 6 for the separated sources. If you are in a hurry, you can upload track 1 straight away. If the edit needs surgery, the stems are there.

To set this up, switch Settings, Output, Output Mode to Advanced, open the Recording tab and tick tracks 1 to 6. Then open Edit, Advanced Audio Properties. Every source gets a row of six checkboxes. Your mic gets 1 and 3, game audio gets 1 and 2, and so on. It looks tedious, but you do it once per scene collection and never again.

Getting Discord off the desktop audio

The step that trips people up is separation itself. The default Desktop Audio source captures everything your PC plays: the game, Discord, a browser tab, the Windows error sound. Ticking it onto track 2 still gives you a mixture, only with a different label.

On Windows 10 version 2004 or later and Windows 11, OBS 28 and newer include an Application Audio Capture source. Add one for the game executable and one for discord.exe, and they arrive in OBS as two clean feeds. Then disable Desktop Audio in the recording, or leave it only on track 1 as a safety net if you are nervous about a program you forgot.

Two Windows settings quietly undo this work, and I have been caught by both:

• The Communications tab in the old Sound control panel defaults to "Reduce the volume of other sounds by 80%" whenever Windows thinks you are on a call. Your game track drops by a huge amount every time Discord opens a voice channel. Set it to "Do nothing".
• Discord has its own Attenuation slider under Voice and Video. Unless you like the ducking, turn it to zero, because it changes the game level that OBS records.

Levels that still make sense once the tracks are apart

Separate tracks are only useful if each one is recorded at a sensible level. A game track that peaks at 0 dBFS will clip, and no editor can unclip it. My rough targets, measured on the OBS mixer meters:

Your mic should peak somewhere around minus 10 dBFS when you speak with normal energy, with loud laughs pushing toward minus 6. I wrote about this more fully in setting microphone gain before a guest arrives, and the same logic applies to a gaming session. Game audio should sit lower than you think, about 12 to 18 dB under the voice in the full mix on track 1. In the separate game track you can record it a little hotter, since you will balance it later anyway.

Set the OBS sample rate to 48 kHz under Settings, Audio, and set every Windows playback and recording device to 48 kHz as well. A Bluetooth headset running at 44.1 kHz while OBS runs at 48 kHz is the most common cause of a voice that slowly slides out of sync with the picture over a long session. I lost most of an evening to that once before realising the headset, not OBS, was the problem.

Recording format and getting the tracks into an editor

Record to MKV. If OBS or the PC crashes, an MKV file is still readable up to the moment of the crash, while a plain MP4 can be lost entirely. Newer OBS builds offer a hybrid MP4 that is safer than the old one, but MKV remains the version I trust on a long session.

After recording, File, Remux Recordings converts the MKV to MP4 in a few seconds without re-encoding, and all six tracks survive the trip. Premiere Pro and DaVinci Resolve both read each track as a separate audio channel or clip. In Resolve, check Clip Attributes, Audio, to see the six channels and assign them. In Premiere, the Modify, Audio Channels dialog does the same job.

If you stream as well as record, remember that a platform stream carries one track only. Twitch's VOD track option in OBS lets you send a second track to the VOD without your music, which is useful, but it is a separate decision from the recording layout above. Keep the two in your head as different outputs. For a wider look at which streaming tools are worth paying for, see picking streaming software without paying for extras.

What separate tracks will not fix

Splitting sources does not clean them. A teammate on a laptop microphone with a fan running will still sound like a laptop microphone with a fan running, just on its own track. Noise suppression has to happen at the source or in the edit, and there is a limit to how much any filter can rescue. Our piece on how voice chat audio gets cleaned up in real time covers what Discord's own processing does before the audio ever reaches you.

It also costs disk space. Six tracks of AAC at 160 kbps add roughly 430 MB per hour, which is small next to the video but not nothing on a session that runs four hours. I keep tracks 5 and 6 switched off when I know there will be no music or alerts.

More recording and editing guides for game footage live in the games section.

One thing to do before your next session

Before you record anything this week, make a five minute test. Set up the six track layout, add Application Audio Capture for the game and for Discord, talk over some gameplay while a friend speaks in a voice channel, then drop the file into your editor and solo each track. If any track contains a sound that should not be there, fix the routing now, while the stakes are a test clip and not the best line of the night.