Tutorials Audio Engineering

Audio Compression Explained for Beginners

By Infinite Creation · Beginner · ~25 min

A 1960s mid-century comic book style editorial illustration showing two audio engineers at a mixing desk, with a loudness meter visible in the frame.

Overview

Compression is the most misunderstood tool in audio production. It doesn't make audio louder. It reduces the dynamic range (the gap between quiet and loud) so the whole performance can sit at a more consistent level. Once you understand the four main controls, you can use compression purposefully rather than by guessing.

What You Need

Any compressor plugin (your DAW's built-in works fine for learning)
A vocal or podcast recording with dynamic variation
Headphones or monitors

Steps

What compression actually does

A compressor is an automatic volume control. When the signal exceeds a set threshold, the compressor turns it down by a set ratio. That's the entire concept. The artistic magic comes from choosing when (threshold), how much (ratio), how fast (attack), and how long (release) to apply that volume reduction. Use our Dynamic Range Compressor tool to see the transfer curve visually.

Set the threshold

The threshold is the level above which compression begins. Lower the threshold until the gain reduction meter is moving on the loudest phrases. For voice: start around −18dB and adjust until you see 4 to 8dB of gain reduction on peaks. If nothing is being compressed, lower the threshold. If it's always compressed, raise it.

Choose your ratio

The ratio determines how aggressively levels above the threshold are reduced. 2:1 is gentle, 2dB above threshold becomes 1dB above. 4:1 is medium, used for most podcast vocals. 8:1+ is heavy limiting. For podcast and VO: 3:1 to 4:1 is the target range. Above 8:1 starts to sound squashed unless that's intentional.

Set attack and release

Attack is how quickly compression kicks in after the signal exceeds the threshold. Slow attack (20 to 50ms) lets the initial transient through before clamping down, preserving the "crack" of consonants. Fast attack (1 to 5ms) catches everything, sounding more controlled but less natural. Release is how quickly the compressor lets go. For voice: attack 10 to 30ms, release 60 to 200ms. A release that's too short causes pumping. Too long causes the compressor to never fully open between words.

Apply make-up gain

Compression turns loud things down, which makes the overall level lower. Make-up gain brings it back up, so the compressed signal matches (or slightly exceeds) the uncompressed level. Without make-up gain, the compressed version sounds quieter and you incorrectly judge it as "worse." Many compressors have an auto make-up gain option. It's fine for learning but manual control is more precise.

Check the gain reduction meter

Watch the gain reduction meter while the audio plays. You want it moving actively on the loudest moments: if it's pinned continuously, your threshold is too low or ratio too high. Aim for the needle swinging between 0 and −8dB on peaks, with quieter sections showing near zero. That's the sign of transparent, working compression.

Pro Tips

The bypass test: toggle the compressor on and off while matching perceived volume with make-up gain. If you can't tell the difference, you've set it well.
Two compressors in series can sound more natural than one heavy one, but only if each stage has its own threshold, attack and release. At the same threshold, 4:1 then 2:1 gives the same curve as a single 8:1.
Audacity's built-in compressor is fine for podcasts. For more control, try the free Airwindows plugins or TDR Kotelnikov.

Tested: the transfer curve, the attack time and the two-compressor tip

We tested step 3 and step 4 and the Knowledge Base example on 2026-09-30 with FFmpeg 8.1.2's acompressor, set to a hard knee, at a -20 dBFS threshold, on steady 1 kHz tones. The columns are the output peak for each input peak. This overlaps the compressor measurements in the guides for de-essing and reverb, and adds the page's own numbers.

Input peak	2:1 (dB out)	4:1 (dB out)	8:1 (dB out)	4:1 then 2:1 (dB out)
-40 dBFS	-40.0	-40.0	-40.0	-40.0
-20 dBFS	-20.0	-20.0	-20.0	-20.0
-16 dBFS	-17.9	-18.9	-19.3	-19.3
-10 dBFS	-14.9	-17.4	-18.6	-18.6
0 dBFS	-9.9	-14.9	-17.3	-17.3

The Knowledge Base example holds to within 0.1 dB. A 4:1 compressor with a -20 dB threshold turned -16 dB into -18.9 dB, the page's -19, and step 3's 2:1 turned -10 dB, 10 dB over the threshold, into -14.9 dB, 5 dB over. 8:1 held -17.3 dB at full scale, which is why "8:1 and above is heavy limiting" is fair: a 20 dB overshoot came out 2.7 dB over.

The tip about two light compressions needed a caveat, which it now carries. At the same threshold, 4:1 followed by 2:1 measured within 0.0 dB of a single 8:1 at every level, because the ratios multiply above the threshold. Two stages can still sound different, since each can have its own threshold, attack and release, and a gentler first stage can react to different peaks, but the transfer curve alone gives no advantage. The page does not say whether the thresholds should differ.

Attack behaves as step 4 says. The first 5 ms of a burst from -60 to -10 dBFS, through an 8:1 compressor with a -30 dBFS threshold, peaked above the settled level by 6.7 dB at 1 ms, 11.6 dB at 5 ms, 15.0 dB at 20 ms, 16.1 dB at 50 ms of attack time, so a slow attack passes more of the opening transient before the gain comes down, which is the "crack" the page describes. Whether that sounds better was not tested.

Validated. Measured 2026-09-30 with FFmpeg 8.1.2 (acompressor, hard knee) and numpy on steady tones and one burst. Compressors differ in detector and knee, and a plugin's curve may not match FFmpeg's. The series result is a transfer curve and says nothing about sound.

Knowledge Base

What You'll Learn

Understand audio compression fundamentally. Learn threshold, ratio, attack, release, and when to use compression on different tracks. Essential for mixing, voiceovers, and podcasts.

What Compression Does

Compressor reduces volume when audio exceeds a threshold. A 4:1 compressor with -20 dB threshold means: if audio is -16 dB (4 dB over threshold), output is only -19 dB (1 dB rise instead of 4 dB).

Key Parameters

Threshold: dB level where compression starts. Lower threshold affects more audio.
Ratio: How much compression (4:1 = moderate, ∞:1 = limiting/hard ceiling).
Attack: How fast compression reacts. Fast (1ms) catches peaks. Slow (100ms) lets transients through.
Release: How fast compression stops. Fast (10ms) releases quickly. Slow (500ms) holds longer.
Makeup Gain: Compensate for compression so output matches input level.

Practical Settings

Vocals: 4:1 ratio, -20 dB threshold, 10ms attack, 100ms release. Glues voice together.
Bass: 4:1 ratio, -15 dB threshold, 20ms attack, 150ms release. Tightens low end.
Drums: 2:1 ratio, -10 dB threshold, 5ms attack, 50ms release. Subtle, natural.
Limiting: ∞:1 ratio, -3 dB threshold (safety ceiling). Prevents clipping.

Where This Fits

Compression is one tool in the mix, and Mixing a Full Song shows where it fits between balance and bus processing.

FAQ

Q: How much compression is too much?
A: 3-6 dB gain reduction sounds natural. 10+ dB is noticeable and requires testing.