How to Record Computer Audio in OBS

Two separate things have to go right before your computer's sound ends up in an OBS recording, and most guides only cover the first one.

The first is capture: some source in OBS has to be listening to your system's output. The second is writing: the audio track that source lands on has to be a track your recording is actually saving. Get the first right and the second wrong and you get the single most common complaint in the OBS forums, which is a mixer full of bouncing green meters and an exported file with silence in it.

This page follows the whole path, from sound leaving an application to sound existing in a file you can hand to someone, and ends with a thirty second test that catches the failure before you record anything long.

The Short Answer

If you want the four steps without the reasoning:

1. Settings → Audio → Desktop Audio, set it to your speakers or headphones instead of Disabled.
2. In Settings → Output, put Output Mode on Advanced, go to Recording, and tick Audio Track 1.
3. Open Advanced Audio Properties from the mixer's right-click menu and confirm Desktop Audio has Track 1 ticked.
4. Record ten seconds, stop, and open the file in something other than OBS.

Step four is not optional. Everything below is why.

Where the Sound Comes From

OBS has three routes to your computer's audio, and which ones you get depends on your operating system more than on your preference.

Recent OBS builds also let a video source carry its own sound, via a checkbox in the properties of Window Capture and Game Capture, which saves adding an audio source at all.

Deciding between these routes is its own question, mostly about what you want kept out of the recording. The desktop audio setup guide works through each option, the notification problem, and the Mac specifics. That page ends where the meter starts moving. This one starts there and follows the sound the rest of the way into a file.

One rule applies regardless of route: if you add a per-application source, set the global Desktop Audio back to Disabled. OBS will happily capture the same sound twice, and the result is a hollow, slightly phased echo that is very hard to diagnose once the take is finished.

Making the Recording Keep It

This is the half that gets skipped, and it is where silent files come from.

Switch to Advanced output mode

Under Settings → Output, move Output Mode off Simple and onto Advanced, then go to the Recording tab. Simple mode hides both the choice of which tracks get written and the bitrate for each one. Advanced costs you nothing for recording; it is the identical encoder with its controls visible.

Tick the tracks you intend to write

Six numbered Audio Track boxes sit in that tab. Whatever you tick here is what the file receives. Sound routed anywhere else is still captured, still mixed and still metered, and then dropped the instant OBS writes to disk.

Tick 1 at minimum. Tick 1 and 2 if you want your computer audio and your voice separable later, which costs almost nothing and is covered in full in how to set up multiple audio tracks in OBS.

Route the source to a ticked track

From the mixer's context menu, open Advanced Audio Properties. Each source gets a row, ending in a grid of track boxes.

Out of the box, every box on every row is ticked. That feels like a safe default and is exactly why nobody suspects routing when a file turns out silent. It stops being safe the first time you untick anything, and plenty of tutorials tell you to.

The rule is short: at least one track ticked in this window must also be ticked in the Recording tab. If Desktop Audio is only on Track 3 and your recording writes Tracks 1 and 2, the file is silent no matter how energetically the meter moves.

Set the audio bitrate

Still in Settings → Output → Recording, bitrate is set per track. The 160 kbps default is fine for speech. If the recording contains music, a product's sound design, or anything you might process further, 192 kbps is a better default, and 320 kbps costs a few megabytes an hour.

This is AAC, which is lossy, so bitrate not spent here is not recoverable later. It is one of the cheapest quality decisions in the whole application.

The Thirty Second Test

Do this once, before the recording that matters.

1. Play something with obvious sound. Any video works.
2. Record for ten seconds, then stop.
3. Find the file. File → Show Recordings opens the folder.
4. Open it in a player that is not OBS.
5. Listen.

If you hear it, the settings will hold and you are finished. If you do not, the table below is your diagnosis, and it cost you thirty seconds instead of an hour.

One detail derails this test regularly: if you enabled several audio tracks, most players only play track 1. Silent playback can simply mean the sound is on track 2 and the player never looked. In VLC, check Audio → Audio Track before concluding anything.

The Meter Moves and the File Is Silent

If the export really is silent and the routing is already correct, the full no-audio checklist works through what is left, in order.

Computer Audio Without Your Voice

A common variation on this task: you want the application's sound and nothing from your microphone. Recording a webinar, capturing a game's audio for a montage, documenting how a product actually sounds.

Do not simply mute the microphone in the mixer and hope. A muted source can still be routed, and a mute is easy to forget to release. Instead:

• In Advanced Audio Properties, untick every track on the microphone row. It is then captured for monitoring and written nowhere.
• Or set Settings → Audio → Mic/Auxiliary Audio to Disabled, which removes it from the session entirely.

The second is cleaner if you never want it. The first is better if you still want to hear yourself while the file stays clean.

If you want the audio and not the picture at all, record normally and extract the audio afterwards, because OBS always writes a video stream. Any remux tool pulls the track out without re-encoding it.

The Container You Record Into

Computer audio survives the recording, and then the file format decides whether it survives everything after it.

• MKV never corrupts on a crash and holds as many tracks as you write. It is also refused by plenty of upload forms and some editors.
• Hybrid MP4, in OBS 30.2 and later, is crash-safe like MKV and opens like MP4, multiple audio tracks included. It is the right default now.
• Plain MP4 writes its index at the end of the file. A crash part way through leaves a file with no index, which usually means the audio and the video are both unrecoverable.

Already sitting on MKV files? Remuxing them to MP4 rewrites the container only, so the audio comes out bit for bit identical.

What All This Machinery Is Actually For

Every step above exists because OBS is a live broadcast mixer that also happens to record. Per-source capture, per-track routing and independent bitrates are exactly the right tools when you are running a stream with a game, a guest, music and alerts, and you need different mixes going to different destinations at the same time.

For one person recording a product walkthrough, it is four settings screens and a test standing between you and a video, and each one is another chance to end up with a silent file.

Cubix Capture is built for that second job. There is no output mode, no track grid and no bitrate decision: it takes system sound and your voice at once, in sync, and puts the saved effort into the picture, moving the frame toward whatever you are clicking on. The file that lands has audio in it because there was never a routing step to get wrong. Windows and macOS, with a free Starter tier if you want to record the same two minutes both ways.

If you are streaming, or you need application audio isolated from your notifications while music keeps playing to your own headphones, OBS is the correct tool and nothing simpler will do that job.

Keep Reading

• Choosing a capture route, covering what each of the three options includes and excludes.
• Splitting voice and system sound into stems, if you want the mix decision left open.
• The Mac path in detail, including systems older than Ventura.
• When the export is silent anyway and the routing already looks right.
• Picking a container, on its own, without the audio question attached.