♫ Digital audio

Audio Compressor: Threshold, Ratio, Attack and Release Explained

How a dynamic range compressor works, what threshold, ratio, attack, release and knee do, and how to set it on voice and podcasts with starting values.

A compressor is a tool that reduces the difference between the loud and quiet parts of a signal. On a voice it makes words more even and easier to follow; on an instrument it can add tightness or energy. It is also one of the hardest effects to hear at first, because it acts on volume over time rather than on tone in an obvious way. Understanding the parameters one by one is the best way to use it with intent.

The principle

When the signal exceeds a certain threshold, the compressor turns it down. How much it turns it down depends on a fixed ratio. Parts below the threshold remain unchanged. The result is a signal with reduced peaks that can be raised overall without exceeding the maximum: as a consequence, the quiet parts become more audible.

Not to be confused with data compression, such as MP3 or FLAC: here we are talking about dynamic range compression, that is about volume.

The parameters

Threshold

The level, in dBFS, above which the compressor starts working. The lower it is, the more of the signal gets compressed.

Ratio

Indicates how much the signal above the threshold is reduced. With a 4:1 ratio, if the signal exceeds the threshold by 8 dB, at the output it will exceed it by only 2 dB.

Ratio Typical effect
1.5:1 - 2:1 Gentle, nearly transparent compression
3:1 - 4:1 Firm control, common on voice
8:1 and above Heavy compression, close to limiting
Infinity:1 Limiter: the signal never exceeds the threshold

Attack

The time the compressor takes to react when the signal crosses the threshold. A fast attack (a few milliseconds or less) also catches the initial transients, making the sound softer but less punchy. A slow attack lets the start of the word or hit through before stepping in, preserving the natural attack.

Release

The time the compressor takes to return to rest when the signal drops below the threshold. Too fast a release creates a "pumping" effect or distortion in the low frequencies; too slow leaves the compressor turned down on the following words too, making everything duller.

Knee

The knee describes how the compressor kicks in around the threshold. With a hard knee the action starts abruptly; with a soft knee the ratio increases gradually, and the effect sounds more natural on voice.

Make-up gain

Compressing lowers the peaks and therefore the overall level. Make-up gain is used to recover it. When comparing the compressed signal with the original, match the levels: otherwise you will only be judging which one is louder.

Reading the reduction meter

Every compressor shows gain reduction, how many decibels it is turning the signal down at that moment. It is the most useful indicator: for spoken voice, a reduction swinging between 3 and 6 dB at the loudest points is already a significant move.

Starting values for spoken voice

Ratio:      3:1
Attack:     10-20 ms
Release:    80-150 ms (or auto)
Knee:       soft
Threshold:  set for 3-6 dB of reduction on peaks

These are indicative values: every voice and every recording needs adjusting. The goal is not a number, but a voice that stays intelligible and even without sounding squashed.

Serial compression

Instead of one compressor working hard, two compressors in series, each working lightly, often sound better: the first, fast, controls sudden peaks; the second, slower, evens out the overall level. Each reduces by 2-3 dB, and the combined result is more transparent.

Common mistakes

  • Over-compressing: the voice loses life, breaths become as loud as words and background noise rises in the pauses.
  • Attack too fast on consonants: the voice becomes muffled and less intelligible.
  • Compressing before cleaning up: the compressor raises the noise too. Noise reduction and the high-pass filter come first.
  • Using the compressor to fix large level differences: if a guest speaks much more quietly, first adjust the level of their segment manually or with automation.

In my work on spoken voice, I focus above all on listening to the pauses and breaths: if they become too prominent, the compression is excessive, even when the meter looks reasonable.

In short

  • A compressor reduces the difference between loud and quiet parts of the signal.
  • Threshold and ratio decide how much to compress; attack and release how it reacts over time.
  • The gain reduction meter is the most useful reference.
  • Two light compressors in series often sound better than one heavy one.
  • Always compare at equal volume and listen to pauses and breaths.
MM
Marco Marzaduri

Marco Marzaduri works in cybersecurity, digital forensics and web development. A computer science graduate, he combines rigorous technical analysis with attention to the people who use technology.

Read also