How does an audio compressor work? Listen and try it yourself

A compressor changes the dynamics of a sound by holding back the louder parts relative to the quieter ones. Threshold sets the level where it kicks in, ratio sets how much the level above that threshold is reduced, and attack and release set how quickly the gain reduction builds up and lets go. Afterward, you can bring the level back up with makeup gain. That's the basis of ordinary downward compression.

The hard part is hearing which control did what. That's why this page has a compressor you can try as you read: play the same loop, change one setting and compare before and after at a similar average level.

Try the compressor while you read

Press Play demo. You'll hear a synthesized loop with drums and bass. A / Before plays the signal without compression; B / After plays it through the compressor. Leave Level-match A/B on for the first exercises.

Listen and try

Your own sound lab.

Open the demo on its own page ↗

Start the sound lab and press Play demo. No sound on iPhone? Turn off Silent mode.

Start the sound lab

Start with the Noticeable preset: threshold −24 dB, ratio 4:1, attack 10 ms and release 150 ms. Switch between Before and After a few times. Listen to the start of each drum hit, its body and the gaps between hits. Then move one control at a time.

Choose Your own audio to try one of your own clips, mono or stereo. The demo accepts files up to 25 MB and two minutes long. The audio file is decoded in your browser, and the demo doesn't send it to Mixanalytic.

The 3D view illustrates the principle. It isn't a model of any real analog compressor. The signal levels and gain reduction it shows, though, come from the actual processing you're hearing.

What does compression do to the sound?

Use a compressor when you want to change how a sound's level moves over time. You might want to hold back a few loud hits, for example, or change the balance between the transient and the rest of a sound. Judge the result in the track where the sound actually has to work.

Start with a concrete problem: “The snare jumps out on some hits” gets you further than “This track needs a compressor.” It also gives you something to check afterward. Is the snare more even now, and has it kept the energy you want?

Dynamic range compression is not the same as file compression. The controls here change how the level of the music moves over time. They don't set an MP3 bitrate or make the audio file smaller. The demo plays back the result; it doesn't export a new master.

The four controls in the demo

Threshold: when does the compressor kick in?

The threshold is the level the compressor reacts to. In the ideal hard-knee model, compression starts when the detected level crosses the threshold. A lower threshold puts more of the same signal above it. With a soft knee, compression comes in more gradually around the threshold. FabFilter's explanation of threshold and knee

Try −12 dB and then −30 dB in the demo without changing the ratio. Watch Gain reduction and listen to how much of the loop is affected. A threshold setting means little until you know the input level: the same value can work hard on a loud clip and barely at all on a quiet one.

In the demo, Volume changes your listening level after the processing. So you can turn it down without changing how hard the compressor works.

Ratio: what does 4:1 mean?

A 4:1 ratio means that, in the ideal static model, a 4 dB rise in input level above the threshold gives a 1 dB rise in output level. The ratio applies to the part above the threshold, not to the signal's entire level. Definition of ratio

Say the threshold is −24 dBFS and the input is −12 dBFS. The input is then 12 dB above the threshold. At 4:1, 12 / 4 = 3 dB remains above the threshold. The output is −24 + 3 = −21 dBFS, which means 9 dB of gain reduction before makeup gain.

The table is a worked example, not a prediction for every moment of the loop. Timing, makeup gain and level matching all affect what the meters show. Use the curve to understand the input/output relationship and A/B to judge the sound.

Attack: how much of the transient gets through?

Attack sets how quickly the compressor increases its gain reduction. Don't read “10 ms” as the compressor waiting ten milliseconds and then suddenly clamping down. Gain reduction builds up gradually over that time. In the Web Audio compressor the demo uses, attack is defined as the time it takes to reduce the gain by 10 dB. Other compressors define and measure attack in other ways. Definition of the demo's attack control

The short burst at the start of a drum hit is called the transient. Listen for whether a shorter attack rounds off the transient and whether a longer one lets more of it through. The change can be easier to hear on drums than on a smooth, sustained sound. FabFilter's guide to the time controls

Keep threshold and ratio at the same values while you compare. Otherwise you'll have to untangle several changes at once.

Release: how does the compressor let go between hits?

Release sets how quickly the compressor backs off its gain reduction. In the demo's Web Audio compressor, it's the time it takes for the gain to come back up by 10 dB. So the value doesn't mean gain reduction always returns to zero exactly that many milliseconds after each hit. Definition of the demo's release control

Compare 50 ms with 500 ms and focus on the gaps between the drum hits. Listen for whether the body of the sound or the background changes level in a way you don't want. No single release value suits all material; judge the recovery against the groove you're working with.

Start with a setting where the recovery is easy to hear, then move the control toward a subtler effect. This is a listening exercise, not a rule for how a finished mix should sound.

Four listening exercises to make sense of the controls

Start with the demo loop so you learn one sound well. Keep level matching on. Switch to your own material once you can describe what a control changes.

1. Hear the difference between less and more compression

Choose Noticeable and listen to After. Compare threshold −12 dB with −30 dB. Leave the other controls where they are.

Then switch between A and B with the threshold at −30 dB. Try to describe the difference without using the words “better” or “worse.” Have the hits become more even? Has their body changed? That description helps you decide whether the change suits the music.

2. Compare a lower and a higher ratio

Set the threshold to −24 dB. Compare ratio 2:1 with 10:1. Leave attack and release unchanged.

Watch the gain reduction meter first, then listen without looking at the number. A bigger number tells you the compressor is turning down more, not whether the result serves your goal. Finish by switching to Before and ask yourself whether you actually prefer either compressed version.

3. Hear what attack does to the drum hits

Use threshold −26 dB, ratio 6:1 and release 150 ms. Switch the attack between 0.5 ms and 60 ms. The extreme values are deliberate for this exercise.

Focus on the very first moment of each hit. Then try a value between the two extremes. You don't have to use either of them in a real mix; the goal is to connect an audible change to the right control.

4. Follow the recovery with release

Keep an attack setting where you can hear the compression, and compare release 50 ms with 500 ms. Listen to how the body and the background come back between hits. Try letting the loop play while you look away from the meters.

Finally, switch between Before and After. If it's hard to remember how the original felt, compare short sections instead of listening to one version for a long time. If you open the demo on its own page, you'll also find a three-step guided listening tour.

How to read the meters without drawing the wrong conclusions

Input and Output show the short-term RMS of the signal before and after the listening path you've selected. RMS is the signal's effective level over the measurement window. The demo measures a mono sum for the level readouts; what you hear stays in stereo. Gain reduction shows how many dB the compressor is turning the signal down while you listen to After. In Before mode, the meter reads zero, because no compression is applied to what you hear. The gain reduction meter rises as soon as the compressor turns the signal down but falls slowly: a fixed time constant of about a third of a second limits how fast it can drop, whatever the release setting. With a short release, it can therefore show more gain reduction between hits than the compressor is actually applying.

Don't read the difference between two RMS numbers as an exact measure of gain reduction. In the demo, makeup gain, level matching and the compressor's timing all affect what you see. Use the gain reduction meter to see how hard the compressor is working on the hits, and Input/Output to see the signal levels.

The dashed threshold line in the history graph is only a reference. The graph shows RMS, so the points where the curves cross the line don't mark exactly when the compressor's detector starts or stops reacting. Nor does the compressor release all its gain reduction the moment a point on the graph drops below the line.

Peak, crest factor and LUFS each measure something different, so it pays to keep them straight. Read more in Mixanalytic's guide to dynamic range, crest factor and LUFS.

Why should you level-match when you compare?

Separate two questions: do you like the change in dynamics, and do you like the change in listening level? Leave Level-match A/B on while you answer the first question. Turn it off if you also want to compare loudness without the demo's extra RMS matching.

The browser's compressor applies automatic makeup gain. On top of that, the demo's level matching adds a fixed gain offset calculated from the clip's RMS before and after processing. The matching doesn't ride the level from hit to hit like an extra automatic volume control. The Web Audio specification describes the compressor's processing.

RMS matching doesn't guarantee identical perceived loudness, and it isn't LUFS matching either. You still need to listen critically, especially when the tone and the transients change. EBU documentation on loudness metering

From exercise to your own mix

Use the same approach on a vocal, a bass or a drum bus: pick one problem, compare before and after, and check the result in the full arrangement. Don't start by copying all four settings from a different sound.

The table suggests a workflow, not a set of ready-made settings. If a single phrase or note needs a precise level change, you can also try volume automation and compare it with compression. The exercise can lead to a useful decision even if you end up not using a compressor at all.

Common mistakes when you're learning compression

You move every control at once. Go back to Noticeable and change one value. It helps to write down one sentence about what you hear before you move on.

You chase a particular gain reduction number. First decide what should sound different. The number can help you find a result again, but it can't choose the result for you.

You compare versions at different levels without noticing. Turn on level matching and redo the comparison. Pick one listening level and stick with it.

You only judge a soloed track. Switch back to the full arrangement. A change you like in solo can come across differently once the other instruments are back in.

You expect every plugin to give the same result. Use the demo to learn the principles. Then check your own compressor's manual for how its timing controls and makeup gain work.

Knee, sidechain and limiter: what doesn't the demo show?

Knee describes the transition around the threshold. A hard knee has a sharper break point in the static curve; a soft knee gives a gradual transition. The demo uses a fixed hard knee, so you can't compare the two with a knee control here. FabFilter's explanation of knee

External sidechain means another signal controls the compressor's gain reduction. A typical example is a compressor on the bass that is triggered by the kick drum. The demo has no external sidechain input. Ableton's manual shows how the sidechain signal and the processed signal are kept separate. Sidechain in Live's Compressor

A limiter is built to keep peaks below a ceiling. An ideal static limiter can be described as having an infinite ratio, but a high ratio on a compressor doesn't automatically give you the same ceiling guarantee or true-peak control. So don't treat a 20:1 ratio in the demo as proof that a master stays below a given peak level. Ableton's Limiter manual

The demo doesn't offer multiband compression, analog coloration or a way to export the processed audio. Use it to explore compression, then do the actual work in your own DAW.

When is it better to keep the original?

If you prefer Before when you compare at a similar average level, keep Before. Don't keep the compression just because the controls are already dialed in.

Go back to the problem you started with. If it's solved but the transient now has the wrong character, look at the attack. If the problem no longer bothers you in the arrangement, try it without processing. A useful comparison can also end with the original winning.

For more on the transient itself, see Mixanalytic's guide to transients and punch. It complements this listening exercise with another way to examine your drum sound.

Frequently asked questions about audio compressors

Does a compressor make audio louder?

Standard downward compression turns the louder parts down relative to the quieter ones. Makeup gain can then raise the whole signal. In the demo, the compressor applies automatic makeup gain, and level matching helps you compare the dynamics at a similar RMS level.

What compressor settings should a beginner start with?

Start with Noticeable in the demo: −24 dB threshold, 4:1 ratio, 10 ms attack and 150 ms release. These values are a starting point for this loop. On your own material, the main thing to work out is how the signal level relates to the threshold.

How do I hear the difference when the compressor barely seems to do anything?

Try a lower threshold and a higher ratio, then switch the attack between two extremes. Compare the same short section in Before and After. Back the effect off once you can recognize what you're changing.

Can I try my own music?

Yes. Choose Your own audio and pick a mono or stereo clip no longer than two minutes and no larger than 25 MB. The file is processed locally in your browser. If the format can't be read, you'll see an error message and can keep using the previous source.

Can I use the demo for mastering?

Use it to learn how the controls affect a clip and to compare settings. It doesn't export a master, and it doesn't check true-peak or loudness targets. Finish and check your deliverables in the setup you use for production.