> ## Content Index
> Fetch the complete content index at: https://obsguides.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# How to record good audio in OBS: levels, filters and tracks
- URL: https://obsguides.com/record-good-audio-obs/
- Published: 2026-11-03T23:15:50.000Z
- Updated: 2026-11-10T14:24:47.000Z
- Description: Set the input level before OBS, use filters for audible problems, and save a full mix alongside separate voice and game tracks. Follow OBS 33 controls and choose AAC or uncompressed PCM audio.
- Author: Ryan Parker
- Tags: Audio

To record good audio in OBS, set a clean microphone level at the device, balance speech above the game, and save a complete mix alongside separate voice and game tracks. Add a filter for a problem you can hear, then check the saved recording. I’d fix the input and routing before increasing the bitrate, because a higher bitrate preserves doubled or distorted sound too.

The controls below follow OBS Studio 33\. Start with the microphone alone; bring in game sound once the voice is clean.

## Give each sound one route into OBS

Open **Settings > Audio**. Keep **Sample Rate** at **48 kHz** and **Channels** at **Stereo** for a normal stereo video project. These are the [OBS 33 defaults](https://github.com/obsproject/obs-studio/blob/33.0.1/frontend/widgets/OBSBasic.cpp?ref=obsguides.com). An existing 44.1 kHz audio workflow can justify keeping 44.1 kHz; changing the number alone won’t clean a noisy microphone.

A mono microphone can sit inside that stereo project. Don’t change the whole project to Mono just to centre one mic.

1. In **Sources**, click **+**, choose **Audio Input Capture**, and select the microphone or interface you intend to use. Give it a recognisable name, such as **Mic**.
2. Check the global microphone assignments in **Settings > Audio**. If the same physical mic is already enabled there, use that existing source or disable its global assignment before adding a separate capture. Keep one route for the device.
3. Speak, pause and tap a key while watching the Mic meter. Confirm that it responds to the intended input before adding filters or game sound.

For **Windows game or application audio**, add **Application Audio Capture (BETA)** and select the application’s window. The [built-in application capture route](https://obsproject.com/kb/application-audio-capture-guide?ref=obsguides.com) supports Windows 10 version 2004 or later and Windows 11\. Disable overlapping **Desktop Audio** capture in Settings > Audio so the same game doesn’t arrive twice. Some applications need a different capture route; a moving mic meter doesn’t establish that the game is being captured.

On **macOS 13 or later**, OBS 30 and later offer **macOS Audio Capture** for the desktop or a selected application. Use the [Mac capture source](https://obsproject.com/kb/macos-desktop-audio-capture-guide?ref=obsguides.com) at this step instead of following Windows device instructions. If macOS Screen Capture already supplies that app’s sound, avoid adding it again through another source.

Want game sound without commentary? Mute or exclude the Mic source while keeping the game source active. If muting the mic also removes the game, those sounds haven’t reached OBS independently yet. Fix that routing before trying to separate them with track checkboxes.

## Set the device level, then balance the mixer

Speak at the level you’ll use during the recording, including the loud phrases. Adjust the microphone or interface’s input gain first. OBS’s [Gain filter guidance](https://obsproject.com/kb/gain-filter?ref=obsguides.com) puts gain at the source before OBS; the filter is there for remaining level adjustment.

If the device is already distorting, lowering its OBS fader only makes the distorted signal quieter. Reduce gain earlier on the path. Leave enough room for your loud voice without driving the input into clipping, then use the mixer fader to place that clean voice against the rest of the recording.

| Sound                            | OBS meter guide                                             | What to confirm in playback                               |
| -------------------------------- | ----------------------------------------------------------- | --------------------------------------------------------- |
| Speech                           | Upper yellow, with louder peaks reaching the lower red area | Quiet words remain intelligible; loud words don’t distort |
| Game or other foreground content | Lower yellow                                                | Effects remain audible without burying speech             |
| Background music and alerts      | Green                                                       | They sit behind the voice instead of competing with it    |

These relative zones come from the [OBS Audio Mixer guide](https://obsproject.com/kb/audio-mixer-guide?ref=obsguides.com). Red means close to clipping, so don’t aim for the far end of it. Meter height also isn’t a complete loudness comparison: a sustained game sound can mask words even when its peaks look similar to the mic’s.

Bring game sound up while speaking, then lower the game fader until every word remains easy to follow in a saved clip. Don’t compensate for a loud game by pushing an already clean microphone toward clipping.

**Voice only in one ear:** a true mono source is automatically heard on both sides. If an interface appears as stereo but only one channel carries the mic, open **Advanced Audio Properties** from the Audio Mixer and enable **Mono** for that source. Keep the game stereo. This corrects the source’s channel layout without collapsing the whole project.

## Choose filters by the defect you hear

Right-click the microphone’s mixer control and open **Filters**. Add one filter, make a short recording, and compare it with that filter disabled. Keep the filter only if it improves the voice without losing words or adding audible pumping.

| Problem                               | Filter to consider                | Its limit                                                                               |
| ------------------------------------- | --------------------------------- | --------------------------------------------------------------------------------------- |
| Steady fan noise under speech         | Noise Suppression                 | It can reduce mild noise; a loud room still reaches the mic                             |
| Noise between phrases                 | Noise Gate, or a gentler Expander | A gate passes background noise whenever your voice opens it                             |
| Loud phrases jump above normal speech | Compressor                        | It reduces dynamics; excessive reduction can make speech pump                           |
| Occasional peaks need a ceiling       | Limiter                           | It can constrain this source after earlier processing; it can’t repair a clipped device |

**Noise Suppression:** start with **RNNoise** for a noisy spoken-voice source. The [OBS suppression filter](https://obsproject.com/kb/noise-suppression-filter?ref=obsguides.com) targets mild background noise such as fans. Speex offers an adjustable suppression level, but stronger suppression can damage the voice too. If consonants sound watery or music loses detail, compare with suppression disabled. Don’t copy the microphone’s speech-cleanup filters onto the game source.

### Set both gate thresholds from your own voice

In **Noise Gate**, work through the fields in the dialog’s order. [OBS’s threshold guidance](https://obsproject.com/kb/noise-gate-filter?ref=obsguides.com) relates the settings to the noise and quiet speech reaching this source:

1. With the gate disabled, note the idle noise level while you aren’t speaking. Set **Close Threshold** a little above that noise so the gate can close between phrases.
2. Set **Open Threshold** below your quietest intended speech but above the noise. It normally sits higher, meaning less negative, than Close Threshold. Enable the gate and confirm that quiet words open it.
3. Check **Attack Time** against the beginning of words. Check **Hold Time** and **Release Time** against pauses and word endings. Change the relevant field when you hear a cut, then record again.

[![OBS Studio Filters for 'Mic' window on Windows 11 with Noise Gate selected: Close Threshold -32.00 dB, Open Threshold -26.00 dB, 25, 200 and 150 ms](https://storage.ghost.io/c/1f/f4/1ff41619-63d0-4562-aeda-c43118aa7da0/content/images/2026/11/03-obs-studio-ui-record-good-audio-obs-0-11c0b3ed85cd.webp)](https://storage.ghost.io/c/1f/f4/1ff41619-63d0-4562-aeda-c43118aa7da0/content/images/2026/11/03-obs-studio-ui-record-good-audio-obs-0-11c0b3ed85cd.webp?ref=obsguides.com) 

Use quiet speech and room noise to set Open Threshold. The values shown are OBS defaults, not a preset for your microphone.

Keyboard taps opening a closed gate point to **Open Threshold**. Raising Close Threshold alone doesn’t stop the opening trigger. If the taps are as loud as your quiet voice, there may be no threshold that rejects one while keeping the other; improve what reaches the mic before making the gate more aggressive.

For a less abrupt fall into silence, an [Expander can replace the gate](https://obsproject.com/kb/expander-filter?ref=obsguides.com). Its ratio reduces quiet sounds instead of imposing the same hard opening behaviour. OBS places this alternative after compression and before a limiter. Recheck its threshold whenever an earlier filter changes the level.

### Control voice peaks without flattening every sentence

A [Compressor starts reducing gain above Threshold](https://obsproject.com/kb/compressor-filter?ref=obsguides.com). Put that threshold where the loud phrases need restraint. Ratio determines the strength: 2:1 is gentler than 6:1\. Adjust Attack and Release while listening for clipped consonants or the level audibly rising and falling around words; use Output Gain only to restore a useful average level afterward.

With a microphone compressor, leave **Sidechain/Ducking Source** unset unless you deliberately want another source to control it. Ducking the game under speech is a separate arrangement: the compressor belongs on the game, with the mic selected as its sidechain source.

If you need several filters, one workable microphone chain is **Noise Suppression → Noise Gate → Compressor → Limiter**, using only the stages your recording needs. That places noise cleanup before gain reduction and the final ceiling after it. Filters process from top to bottom in the list. An Expander alternative goes after the Compressor, ahead of the Limiter.

Keep the [Limiter last](https://obsproject.com/kb/limiter-filter?ref=obsguides.com). Its default Threshold is −6 dB with a 60 ms Release; those are starting values to inspect, not a loudness target for every source. A mic limiter also doesn’t constrain the combined sum of mic, game and music, so check the complete mix.

## Save a full mix and isolated recording tracks

I’d keep a complete mix on track 1 because an ordinary video player usually plays one audio track at a time. Separate voice and game tracks then give the editor room to rebalance them. OBS’s [multiple-track workflow](https://obsproject.com/kb/multiple-audio-track-recording-guide?ref=obsguides.com) requires both source assignments and enabled recording tracks.

Open **Advanced Audio Properties** from the Audio Mixer. For a two-source recording, use this example:

| Source | Track 1: complete mix | Track 2: voice | Track 3: game |
| ------ | --------------------- | -------------- | ------------- |
| Mic    | Checked               | Checked        | Unchecked     |
| Game   | Checked               | Unchecked      | Checked       |

Then open **Settings > Output**, set **Output Mode** to **Advanced**, and open **Recording**. Keep **Type** at **Standard**. Under **Audio Track**, enable **1, 2 and 3**, then apply the settings. Assigning the mic to track 2 won’t save track 2 if that output checkbox is off.

[![OBS Studio Advanced Audio Properties on Windows 11 with Mic ticked on tracks 1 and 2 and Game ticked on tracks 1 and 3](https://storage.ghost.io/c/1f/f4/1ff41619-63d0-4562-aeda-c43118aa7da0/content/images/2026/11/03-obs-studio-ui-record-good-audio-obs-1-9699aac5d647.webp)](https://storage.ghost.io/c/1f/f4/1ff41619-63d0-4562-aeda-c43118aa7da0/content/images/2026/11/03-obs-studio-ui-record-good-audio-obs-1-9699aac5d647.webp?ref=obsguides.com) 

Keep both sources on track 1 for playback, mic alone on 2 and game alone on 3\. Enable the same three tracks in Recording.

Keep unrelated sources off the isolated tracks. When editing, use either the complete mix or the isolated tracks for that passage; playing both together doubles the material. If the editor exposes only one track, inspect its import settings and container/codec support before concluding that OBS failed to record the others.

## Choose compact AAC or an uncompressed editing track

For a straightforward local stereo recording, choose an **AAC** audio encoder with a high available bitrate. In Advanced Output, the **Audio** tab holds each track’s **Audio Bitrate**. I’d use **320 kbps** when small files and broad compatibility matter: it is the upper non-surround bitrate allowed by the [OBS 33 track controls](https://github.com/obsproject/obs-studio/blob/33.0.1/frontend/settings/OBSBasicSettings.cpp?ref=obsguides.com#L2197), and [AAC has broad editing support](https://obsproject.com/kb/audio-video-formats-guide?ref=obsguides.com). Encoder availability determines the choices you see.

For audio that will undergo substantial editing, you can avoid lossy output compression without making the video lossless:

1. In **Output > Recording**, select **Matroska Video (.mkv)** as the Recording Format.
2. Choose **FFmpeg PCM (24-bit)** as the Audio Encoder. Keep a suitable video encoder selected separately.
3. Keep the recording tracks you need enabled. Make a short file and confirm that your editor imports the picture and every required audio track.

[![OBS Studio Settings, Output in Advanced mode, Recording tab on Windows 11: Matroska Video (.mkv), FFmpeg PCM (24-bit) and Audio Track 1 to 3 ticked](https://storage.ghost.io/c/1f/f4/1ff41619-63d0-4562-aeda-c43118aa7da0/content/images/2026/11/03-obs-studio-ui-record-good-audio-obs-2-d3a999e56f3c.webp)](https://storage.ghost.io/c/1f/f4/1ff41619-63d0-4562-aeda-c43118aa7da0/content/images/2026/11/03-obs-studio-ui-record-good-audio-obs-2-d3a999e56f3c.webp?ref=obsguides.com) 

Choose the recording container and audio encoder separately. PCM audio does not require lossless video.

[FFmpeg PCM (24-bit)](https://github.com/obsproject/obs-studio/blob/33.0.1/plugins/obs-ffmpeg/data/locale/en-US.ini?ref=obsguides.com) is the exact built-in encoder label. PCM is uncompressed and takes more storage than AAC. It preserves the signal presented to that encoder at the selected precision; it doesn’t undo filters, recover a clipped mic or restore detail already lost in the input.

MKV accepts the codec combinations OBS supports and tolerates an interrupted recording, but editor support varies. Remuxing changes the container without re-encoding; it doesn’t make an unsupported audio codec compatible with every editor. Verify the destination workflow before a long session. Hybrid MP4 and Hybrid MOV are other resilient recording options, so “all MP4 recordings are unsafe” is an outdated blanket rule.

Don’t choose Simple mode’s **Lossless Quality, Tremendously Large File Size** just to improve audio. That [whole-recording preset uses AVI, UTVideo and 16-bit PCM](https://github.com/obsproject/obs-studio/blob/33.0.1/frontend/utility/SimpleOutput.cpp?ref=obsguides.com), changing the video path too. A high-bitrate AAC or separate PCM audio encoder is the more focused choice.

This setup saves sound inside a video recording. A standalone WAV or other audio-only export is a different output task; an empty scene by itself doesn’t change the recording into an audio file.

## Keep listening and recording as separate checks

In OBS 33, **Advanced Audio Properties > Audio Monitoring** has an **Enable** checkbox for each source. Choose the listening destination under **Settings > Audio > Advanced > Monitoring Device**. Don’t follow an older tutorial’s Monitor Only or Monitor and Output dropdown instructions.

Use headphones and avoid [sending monitored sound](https://obsguides.com/stream-obs-to-discord/) into a desktop endpoint that OBS captures again. If you already hear the game directly, enabling OBS monitoring can add another copy in your headphones. Disable the extra listening path while keeping the game assigned to its recording tracks.

A moving meter proves that sound reached the source’s mixer path. Hearing it in headphones proves that your listening route works. The saved file is what establishes whether the selected tracks contain it.

## Listen to a private rehearsal before the full session

Record about 30 seconds: a quiet phrase, normal speech, a loud phrase, a pause with the usual room noise, then speech over the loudest game section you expect. [Find the saved recording](https://obsguides.com/where-obs-saves-recordings/) through **File > Show Recordings** and listen through headphones at an ordinary playback volume.

Check that track 1 contains the complete mix, both ears carry the intended voice, and quiet beginnings and endings survive the filters. In the editor, listen to tracks 2 and 3 individually to confirm the separation. If one check fails, change the control at that point on the path and record another short clip before committing to the full session.