You searched for how to set a compressor, so the situation is probably this: a plugin with six knobs is open on a channel, three tutorials gave three different sets of numbers, and the sound either does not change or gets worse. The fix is a simpler mental model. A compressor answers exactly two questions: how fast it grabs the signal and how fast it lets go. Attack and release are the sound of the compressor. Ratio and threshold only set how much of that sound you get. Hold onto that and the rest of this guide is just numbers.
How an audio compressor works, in plain terms
Before the knobs, the concept. Here is how a compressor works: you set a threshold, a level in dB, and anything that crosses it gets turned down. The ratio sets how hard. At 4:1 a signal that goes 4 dB over the threshold comes out only 1 dB over, so the loud parts get pulled toward the quiet parts and the whole signal sits in a tighter window. That is the entire job of an audio compressor: it makes the dynamic range smaller by reducing gain on whatever passes the threshold. Attack and release decide the timing of that gain reduction, how fast it clamps down when the signal goes over and how fast it lets go when the signal drops back under. Everything below is just choosing those four values for a given source. If threshold, ratio, or gain reduction still feel abstract, the glossary defines each one in a couple of sentences.
Compressor attack decides the transient: listen to the first 50 ms
Every percussive sound has two parts: a transient, the initial spike that reads as punch and click, and a body, the sustain that reads as weight and tone. The attack time decides which part the compressor touches. A slow attack of 10 to 30 ms lets the transient pass untouched and then squeezes the body behind it. The spike now stands taller above the sustain, so the sound gets punchier. A fast attack under 5 ms clamps down on the transient itself: thicker and more controlled, but duller, because the part your ear uses to locate the hit is gone.
Neither setting is wrong. The test is always the same: loop the sound and listen only to the first 50 ms of each hit. Sweep the attack from 30 ms down toward 0 and you will hear the click survive, then soften, then disappear. Stop while the click is intact if the sound needs to cut through, like a snare in a busy drop, and keep going until the spike is gone if it needs to sit still, like a slappy bass.
Compressor release decides the groove: time it to your tempo
Release controls how fast the compressor returns to zero gain reduction after the signal drops back under the threshold. That timing is a rhythmic decision. If the gain reduction meter is back at zero before the next hit lands, every hit gets the same treatment and the compression sounds clean and invisible. If the compressor is still recovering when the next hit arrives, each hit rides on the tail of the last one and you get pumping: the breathing, glued movement you hear on most electronic drum buses. Clean and pumping are both valid. Choosing between them by accident is the mistake.
Time the release to your tempo instead of guessing. At 128 BPM a quarter note lasts 469 ms and an eighth note lasts 234 ms, so a release around 200 ms on a four-on-the-floor drum bus recovers just before each kick and a release around 400 ms pumps audibly through the bar. For sustained material like pads and vocals, switch on auto-release and let the compressor adapt. And if the pumping you want is the kick-ducking-the-bass kind, that is a separate technique with its own numbers, covered in the guide to sidechain compression settings.
Ratio and threshold are only the amount: aim for 2 to 4 dB
Once attack and release define the character, ratio and threshold decide how much of it you apply. Read the ratio as intensity: 2:1 is gentle glue that the ear barely registers as compression, 4:1 is working compression that audibly reshapes the sound, and 8:1 and above is limiting territory for taming wild peaks. Then set the threshold by watching the gain reduction meter, not the threshold number itself. On most material, pull the threshold down until you see 2 to 4 dB of gain reduction on the loud parts. On drums you deliberately want squashed, 4 to 8 dB. The threshold value that gets you there changes with every source, which is exactly why tutorials quoting threshold numbers never transfer to your session.
How to set a compressor per source: the starting numbers
These are starting points. Dial them in, then run the 50 ms transient test and check the gain reduction meter recovers in time.
Drum bus
Attack 30 ms so every transient passes clean, release timed to the gap between hits, usually 100 to 200 ms, ratio 2:1, and 2 to 4 dB of gain reduction on the loudest hits. This is the glue setting: the kit sounds like one instrument instead of six channels.
Kick and snare individually
Decide the job first. Taming an inconsistent or spiky hit: attack 5 to 10 ms so the compressor catches the front. Adding punch to a soft hit: attack 20 to 30 ms so the transient jumps out before the body gets squeezed. Same compressor, opposite results, and the attack knob is the only difference.
Bass
Bass wants consistency, not punch. Attack 1 to 5 ms, release 100 to 200 ms, ratio 4:1, and 3 to 6 dB of gain reduction. The goal is a note level that barely moves, because an even bass is what lets the low end sit at one fader position for the whole track.
Synths and pads
Slow everything down. Attack 30 ms or more, auto-release or 300 ms plus, ratio 2:1, and only 1 to 3 dB of gain reduction. Pads have no transients to shape, so all you want is a gentle leveling that stops the swells from eating headroom.
Vocals and leads
Ratio 4:1, attack 5 to 15 ms so consonants keep some edge, release 60 to 150 ms so the compressor recovers between words, and 3 to 6 dB of gain reduction on the loud phrases. If one compressor cannot smooth the part without sounding squeezed, that is your cue for the two-in-series trick below.
Once the sources are dialed in, check the result against the genre. Upload a bounce to TrackSensei and read the dynamics and punch sections: it measures your transients and dynamic range against released tracks in your genre, so you know whether the drum bus is glued or just flattened.
Match the makeup gain or you are only hearing louder
Compression turns peaks down, so a compressed signal is quieter, and a quieter signal always loses an A/B test. That single fact explains most bad compression decisions. Set the makeup gain so the compressed signal matches the bypassed level, then toggle bypass and compare at equal loudness. Now you are judging the shape of the sound instead of the volume. And apply the honest rule: if you cannot hear a difference beyond louder at matched levels, the compressor is not doing a job. Lower the ratio or remove it. Where compression sits in the larger order of operations is covered in the general mixing guide.
The three compression mistakes beginners repeat
First, compressing because the tutorial said so. A compressor showing 0.5 dB of gain reduction is doing nothing. Either pull the threshold down until the meter shows 2 dB or more, or delete it. Second, setting the attack to 0 ms by default. That kills every transient on the channel, and a mix full of zero-attack compressors is the flat, lifeless sound every beginner tutorial warns about. Third, asking one compressor to do two jobs. A vocal that needs both peak control and overall leveling sounds strangled through one compressor doing 8 dB. Use two in series, each doing 2 to 3 dB: a faster one catching peaks, a slower one riding the level. Two gentle stages always sound more natural than one aggressive one. And if you suspect your mix already got the aggressive treatment, AI feedback on your mix will show whether the punch survived.
The five-minute routine for any compressor
Run this on every source. Pick the character first: slow attack for punch, fast attack for control. Time the release to the tempo, or auto for sustained material. Set the ratio for the job, 2:1 glue, 4:1 working, 8:1 limiting. Pull the threshold down to 2 to 4 dB of gain reduction. Match the makeup gain and A/B at equal loudness. The whole pass takes five minutes, and four of those five decisions are about time, not amount.
The part you cannot verify by ear alone is whether the result lands where your genre lives, because punch and dynamic range are exactly the things a long session deafens you to. If you want numbers instead of a tired ear, upload your track to TrackSensei and check the dynamics section. It tells you whether your transients and crest factor match released tracks in your genre, and points at the channel where a compressor went one stage too far.