A Compression Guide for Natural-Sounding Vocal Mixes
A practical framework for setting threshold, ratio, attack and release by logic instead of by ear alone — how to get a vocal that's controlled without sounding squashed.
The clearest sign of a mishandled compressor: the vocal is either still jumping around, or it's been flattened into a lifeless line. The difference between the two usually doesn't come from one parameter, but from how four parameters talk to each other. In this piece I'm explaining how to set a compressor by understanding what the signal is doing, not by habit.
Threshold: What Are You Catching?
Before touching the threshold, identify the loudest 3-4 words by ear alone, without watching the gain-reduction meter. Set the threshold so the compressor only catches those peaks. If the whole vocal sits in constant gain reduction, that's usually a sign the threshold is set too low — and that's the first thing to kill natural dynamics.
A practical test for finding the right threshold: bypass the compressor and listen to the vocal, then engage it and listen again — if the difference feels like 'a completely different performance' rather than 'slightly more controlled', the threshold is probably set too aggressively. Good compression does its job without announcing itself to the listener.
Ratio: Control or Effect?
- ―2:1 – 3:1: transparent control close to natural speech — a good starting point for podcasts and dubbing.
- ―4:1 – 6:1: noticeable but still musical compression, common in pop/commercial music.
- ―8:1 and above: you're in limiting territory now — a deliberate effect choice, not a default.
Attack and Release: The Pair That Shapes the Envelope
Opening up the attack lets a word's initial transient (especially hard consonants like 'p', 't', 'k') pass through uncompressed, preserving clarity. Set the attack too fast and the start of the word gets squashed too, leaving a dull, recessed vocal. Release determines how long it takes to recover before the next word — too slow and the compressor 'pumps' from one word to the next; too fast and background noise starts breathing in and out audibly.
The interaction between these two parameters produces results neither one explains alone: a fast attack combined with a slow release grabs the start of a word and doesn't let go for a long time, creating a vocal that feels like it 'can't breathe'. A slow attack with a fast release lets transients through and snaps everything else back quickly — usually a livelier but harder-to-control combination.
Why Serial Compression Sounds More Natural Than a Single Compressor
Trying to pull 10 dB of gain reduction out of a single compressor almost always leaves an audible artefact. Split the work across two compressors instead: let a light, fast one (2:1, medium attack) tame the overall dynamics first, then a slower, more characterful one (opto or VCA-style) add colour and consistency. With each working around 2-4 dB, you get the same total control far more transparently.
This is where the character of the second compressor (opto, VCA, FET, vari-mu) comes into play: since the first compressor has already smoothed out the general dynamics, you can start thinking of the second one not as a technical fix, but as a tone-coloring tool. That's why so many well-known vocal chains stack two different brands or types of compressor in series — each one is doing a different job.
Keep De-Essing Separate From Compression
Sibilance ('s', 'sh' sounds) isn't a problem the general compressor should be solving — it's frequency-specific. Setting the main compressor aggressively enough to tame sibilance dulls the whole vocal. Instead, place a dedicated de-esser at the end of the chain, after makeup gain, targeting only the 5-8 kHz band.
The frequency a de-esser should target varies from voice to voice — some vocals concentrate sibilance around 6 kHz, others push closer to 8-9 kHz. That's why, instead of a fixed 'general-purpose' setting, looking at a spectrum analyzer to see exactly where the 's' sounds peak and aiming the de-esser there produces a far more transparent result.
Makeup Gain: The Last Step, Not the First
When A/B-ing, the ear always perceives the louder signal as 'better' — that's an illusion. Set makeup gain so the compressed and uncompressed signal sit at the same perceived loudness (use a loudness meter if needed); only then can you honestly judge whether the compressor is actually helping.
This step is also the only honest way to judge how much work the compressor is actually doing: if, after setting makeup gain to match perceived levels, the compressed signal still sounds noticeably 'better', that's a real improvement — not just a loudness illusion.