Skip to content
streamneo.
Setup Guides14 min read

How to Set Up Microphone Filters for Screen Recording

Set up OBS microphone filters by problem: test noise suppression, gates, expanders, compression and limiting without relying on a universal preset.

sn.
StreamNeoPublished 4 October 2026
Worth sharing?

Start by adding filters to the microphone source in your recording software, then adjust only the filter that addresses the problem you can hear. For most screen recordings, a short test recording is more useful than copying a complete chain of settings from somebody else.

Noise suppression can help with mild, steady sounds such as a computer fan. A gate or expander deals with unwanted sound during pauses, while compression controls unusually loud speech and a limiter caps the final peaks. None of these can reliably repair a loud room, a poor microphone position or severe clipping.

Add filters to the microphone source in OBS

In OBS Studio, filters belong to an audio source rather than to the whole recording. This distinction matters if your scene contains a microphone, desktop audio, a browser source and perhaps a capture card. A filter added to the microphone source changes the voice input, not every sound in the recording.

In the Audio Mixer, find the microphone or auxiliary audio source you want to change. Select the three-dot menu beside it, choose Filters, and use the plus button in the Audio Filters panel to add a filter. Give filters useful names if you expect to revisit the setup, such as “fan suppression” or “voice limiter”. OBS documents the filter list and its basic controls in the official OBS Filters Guide.

The exact labels and layout can vary with the OBS version and the source you selected. Before changing anything, confirm that the mixer meter moves when you speak and that you have selected the intended microphone. If the wrong source is active, carefully tuned filters will appear to do nothing.

You do not need every available filter. Start with the problem that is most obvious in an unprocessed recording. If the room is quiet but your voice jumps sharply when you become animated, compression may be relevant while suppression is not. If the voice is clear but the fan remains audible during pauses, suppression or an expander may be worth testing.

Keep the unfiltered recording available while you work. In OBS, you can temporarily disable a filter with its visibility or enable control, depending on the interface version. Comparing the processed and unprocessed takes at the same listening volume helps you notice when a filter is making the voice less natural.

Record a short audio test before tuning

Make a short test recording before changing several controls together. Speak in the same position and at roughly the same distance you will use for the real screen recording. Include a normal sentence, a quieter sentence, a deliberate pause, and a louder phrase. If your recording includes keyboard sounds, a click or a notification, include those only if they are part of the situation you are trying to manage.

Listen to the file through headphones, then through the speakers you normally use for editing. Headphones reveal hiss, room tone and clipped consonants more clearly, while speakers help you judge whether the voice still sounds natural at ordinary playback volume. Do not judge only by the OBS meter. The meter shows level, but it cannot tell you whether suppression has made the voice metallic or whether a gate has cut off the end of a word.

Change one control at a time and record another short sample. Write down what you changed, even if the adjustment seems small. A simple note such as “stronger suppression made the ‘s’ sound rough” gives you a way back when several experiments begin to blur together.

Use the same test phrase after each change. A filter can sound acceptable on a single sentence and reveal problems when you pause, start speaking quietly or raise your voice. The most useful test is the one that resembles the final recording, not a deliberately polished microphone demonstration.

Check the beginning and end of words. A gate or expander may reduce room noise successfully but remove a quiet first syllable. Compression may make the average voice level steadier but bring breaths or keyboard noise forward. A limiter may stop a peak while leaving the rest of the recording too quiet. These are listening decisions, not signs that one setting is automatically correct.

If your recording contains desktop audio, listen to the microphone and desktop audio separately as well as together. A voice that sounds clear on its own can still be covered by music or a browser video. For a continuous stream or long recording, also test a few minutes rather than relying only on a short clip. Problems such as pumping and listener fatigue can become more obvious with time.

Use noise suppression for mild, steady noise

Noise suppression is aimed at a fairly specific problem: a continuous or repeated background sound that sits underneath the voice. The OBS Project describes its Noise Suppression filter as suitable for mild background or white noise, including computer-fan noise, but notes that it is generally not effective when the room itself is loud. The OBS Noise Suppression documentation is the reference for the current filter behaviour.

Add the filter to the microphone source and make a recording with it disabled. Then enable it and compare the same phrase. If the fan or light hiss is less distracting and the voice remains clear, the filter is addressing the right problem. If the room contains conversation, traffic, a loud appliance or intermittent bangs, stronger suppression is unlikely to turn it into a clean voice track.

OBS describes RNNoise as the higher-quality method in the filter and notes that it uses more CPU than Speex. Speex provides a suppression-level control, while RNNoise and Speex may present different choices depending on the OBS version. The choice is therefore a trade-off between the type of noise, the available processing capacity and the sound of the result.

Start gently. With Speex, increasing the suppression level makes the filter act more strongly, but excessive suppression can distort the voice. You may hear watery, metallic or chirping artefacts, particularly on consonants and quiet speech. If that happens, reduce the strength or disable suppression and address the sound at its source by moving the microphone, lowering the fan noise or closing the noisy application.

Do not use suppression to compensate for a microphone that is too far away. A closer, sensible microphone position often gives the voice more level relative to the room, although it can introduce plosives, handling noise or a more prominent breath sound. Those are separate problems and should be tested separately.

Noise suppression can also interact with other filters. If a gate follows it, the gate may see a quieter background and close more readily. That can be useful, but it can also make the opening and closing more abrupt. Compare the complete chain, not just each filter in isolation.

Use a gate or expander for silence

A gate is useful when the main problem is noise during the moments when you are not speaking. It reduces or cuts the microphone signal below a threshold. When your voice crosses the opening threshold, the signal is allowed through. This can make pauses quieter, but a gate does not distinguish perfectly between speech and every other sound.

OBS advises setting the close threshold above the background-noise level and the open threshold slightly below the voice input. Its documented software defaults include a close threshold of -32 dB, an open threshold of -26 dB, an attack of 25 ms, a hold of 200 ms and a release of 150 ms. Treat those as defaults in the software, not as a universal voice preset. The OBS Noise Gate guide explains the controls.

First record a pause without speaking and note where the background sits on the meter. Then speak quietly, normally and loudly. The open threshold needs to admit the quietest speech you want to keep, while the close threshold needs to prevent the ordinary room noise from remaining open. There may not be a large gap between those two levels, especially with a distant microphone.

Listen for chopped words. If the first consonant disappears, the opening threshold may be too high, the attack may be too slow or the microphone level may be too low. If the last part of a word vanishes during a pause, the close threshold or release behaviour may be too aggressive. A longer hold or release can sound more natural, but it may allow more room noise through.

An expander is an alternative when a hard gate sounds abrupt. It reduces quiet signals rather than applying the same cut-off behaviour, and its adjustable ratio can make the transition smoother. OBS recommends RMS detection for its expander and describes a ratio of 4:1 as a possible balance, not a setting that suits every microphone or room. Its Expander documentation gives the relevant controls and terms.

Choose a gate when a clean, quiet pause is more important and your speech level is reliably separated from the room. Choose an expander when you want the noise reduced but want quiet speech tails and pauses to remain more natural. In either case, tune the threshold from your actual recording rather than from a screenshot of somebody else’s settings.

A gate or expander will not stop a chair scrape, a door closing or a cough that is louder than the threshold. It may also open for keyboard taps or desk vibrations. If those sounds matter, deal with their cause or record them separately rather than expecting a threshold control to classify them correctly.

Control loud peaks with compression

Compression addresses level variation. It reduces the difference between quieter and louder parts of the microphone signal, particularly when you speak more loudly for emphasis. This can make a screen-recording voice easier to follow, but it does not remove background noise and it does not repair clipping that has already occurred.

OBS describes its compressor as reducing louder microphone peaks and says it is typically placed at or near the beginning of the filter chain. The documented default threshold is -18 dB and the documented default ratio is 10:1. Those are product defaults, not evidence that they are right for your voice. Use them only as software starting references, then test the result using the phrases you will actually record. See the OBS Compressor guide for the control descriptions.

Set the threshold so the compressor responds to the loud parts you want to control rather than constantly working on ordinary speech. If it is active nearly all the time, the recording may sound flattened or may bring up breaths and room noise. If it never responds, it is not addressing the peaks that prompted you to add it.

Listen for pumping. Pumping sounds like the level moving up and down around speech, often because the compressor is reacting strongly to a voice that changes quickly. Also listen for a dull or pressed quality. A compressor can make quiet words more prominent after the loud parts are reduced, but that benefit disappears if the processed voice no longer sounds like you.

After compression, check the output level. Reducing peaks can make the recording seem quieter even when it is easier to hear. Do not immediately compensate by pushing every control upwards. Compare the unprocessed and processed recordings at a similar playback level, then check the loudest phrase for distortion.

OBS also documents sidechain ducking. In that arrangement, a compressor on desktop audio uses the microphone as its sidechain source, lowering the desktop sound while you speak. This is separate from compressing the microphone itself. It can help a tutorial with music or application audio, but test speech and desktop audio together so that the background does not disappear too sharply whenever you say a short word.

Place a limiter last and recheck

A limiter is a final ceiling. It prevents the output from exceeding its threshold, which makes it useful for catching a sudden loud word or an unexpected peak after other processing. It is not a substitute for a sensible microphone level and it should not be used to make an excessively quiet recording loud.

OBS says the limiter should be the last filter in the chain. Its documented default threshold is -6 dB with a 60 ms release. These are starting values supplied by the software rather than a guarantee for every microphone, recording format or platform. The OBS Limiter guide describes the filter and its controls.

Place the limiter after any suppression, gate or expander and compressor you decide to use. Then repeat the loudest phrase in your test recording. Watch whether the limiter is responding occasionally to genuine peaks or working continuously. A limiter that is constantly active can make speech feel constrained and may hide the fact that the input level is set too high.

Listen to the release after a peak. If the level returns abruptly, the result may sound unnatural. If it returns too slowly, the next words may remain reduced. The right balance depends on how quickly you speak and how much level movement your recording can tolerate.

A limiter cannot undo clipping in the microphone input or a distorted preamp. If the source is already clipping before it reaches the filter, lower the microphone gain or move farther from the microphone and record again. Watch the input meter while speaking at the loudest level you expect to use.

Recheck the whole chain after putting the limiter last. Suppression may alter the noise floor, a gate may remove pauses, compression may raise quieter details and the limiter may catch the remaining peaks. A chain that sounds sensible one filter at a time can still sound artificial when all the controls are active.

Choose the smallest useful chain

A practical chain is built from problems, not from a fixed recipe. One person may need only a compressor and limiter. Another may need mild suppression because of a fan, an expander for pauses and a limiter for occasional peaks. A third person may need no filter until the microphone position and room noise have been improved.

What you hear Filter worth testing What to listen for When to stop or change approach
A mild, steady fan or hiss Noise suppression Less background without metallic or watery speech Reduce it if consonants become distorted, or address the sound physically
Room noise mainly during pauses Gate Quieter silence without chopped words Lower the opening threshold or try an expander if quiet speech is lost
A smoother reduction during pauses Expander Natural speech tails with less low-level noise Re-tune the threshold if quiet words are reduced too much
Occasional loud speech peaks Compressor More even speech without pumping Reduce the action if the voice sounds pressed or room noise rises
Sudden final peaks Limiter, last Peaks are caught without constant level reduction Lower the input level if it is working continuously

For an ordinary tutorial, begin with a short test containing normal speech, quiet speech and a pause. Add one filter, listen, and keep it only if it solves a problem without creating a more distracting one. Save a copy of the working profile before experimenting further.

If you record devotional audio, study sessions, local news loops or long demonstrations, monitor for fatigue as well as obvious defects. A voice that sounds impressive for ten seconds may become tiring over an hour if compression is too strong or the gate opens and closes repeatedly. Longer test recordings are especially important when the final content is intended to run continuously.

Keep desktop audio separate where possible. Applying voice processing to a mixed source makes it harder to control speech without also changing music, application sounds or notifications. If your recording method combines sources before they reach OBS, check whether the filter is acting on the microphone alone or on the combined signal.

If the recording must later become part of a continuous YouTube broadcast, solve the source audio before making the long file. A cloud-based service can keep a prepared video running with your computer switched off, but it cannot make a badly recorded microphone track clear. For the wider operational question, compare the approach in how to start a 24/7 YouTube live stream from pre-recorded videos in India. If your current setup depends on a local computer, the guide to alternatives to keeping a computer on for a 24/7 YouTube stream covers a separate decision from microphone processing.

When a long broadcast stops, audio filters are not usually the first place to investigate. Check the recording source, OBS status and the connection path, then use the fix order for a stream stuck on Starting Soon or Waiting for Data. For an outage affecting a running channel, the recovery guide for a YouTube 24/7 stream after an internet outage in India is more relevant than changing the microphone chain.

For a prepared file that needs to play continuously, StreamNeo removes the need to leave your recording computer switched on after the file and YouTube connection are ready, while the audio quality still depends on the recording you supplied.

Before committing, compare the operating options on the pricing page. When the file and channel are ready, start free — 24-hour trial, no card.

FAQ

Should I use every microphone filter in OBS?

No. Add a filter only when it addresses a problem you can hear in a test recording. A smaller chain is often easier to understand and less likely to create pumping, chopped words or distorted speech.

Can noise suppression remove all background noise?

No. OBS describes noise suppression as suitable for mild background or white noise, not as a complete solution for a loud room. Reduce the sound at its source where possible, then use a short recording to check whether suppression improves the result without damaging the voice.

Should a gate come before or after compression?

There is no universal order that suits every recording, because the filters respond to the signal they receive. Tune the gate or expander so it does not remove quiet words, tune compression for loud peaks, and listen to the complete chain rather than relying on a fixed preset.

Why does my microphone still sound distorted with a limiter?

A limiter can cap later peaks, but it cannot repair clipping that happened at the microphone or input stage. Lower the input level, repeat the loudest part of your test and confirm that the source is clean before the limiter receives it.

YOU’VE REACHED THE END

Keep the ideas coming.

More guides, useful tools and a little help for your next broadcast.

Back to the journal ↗
YOUR NEXT READ

A little more to explore.

More Setup Guides guides ↗ · All topics ↗