Compressing vocals naturally, step by step
One compressor doing all the work is audible. Two doing half each are not. Here is the chain I start from almost every time.
The vocal is the most dynamic element in almost every track. A whisper and a full note can be more than 20 dB apart, and even a very steady rap has hard consonants and words that jump out. The job of compression is to deliver every word to the listener without the work being audible.
The secret is not the right compressor but the division of labour: first you level by hand, then a fast compressor catches the peaks, then a slow one adds cohesion. Each does a little, and the result sounds natural.
The chain, step by step
Level with clip gain first
Listen to the track and use clip gain to lower the phrases or syllables that jump out and raise the ones that disappear. A few minutes are enough. The compressor works much better on a signal that is already steady, instead of swinging from 0 to 12 dB of gain reduction within the same phrase.
Clean up with a high-pass filter and subtractive EQ
A high-pass between 70 and 100 Hz for male vocals, between 100 and 150 Hz for female vocals, removes rumble and handling noise that would make the compressor react for nothing. Make sure you do not thin the voice out. If it sounds “boxy”, a narrow cut between 250 and 500 Hz is usually enough.
First compressor: fast, on the peaks only
A FET or VCA-style compressor with a ratio between 4:1 and 8:1, fast attack (under a millisecond on a 1176-type FET, 1 to 5 ms on a VCA) and release between 50 and 100 ms. Set the threshold so it works only on the loudest parts, with 3–6 dB of reduction on the peaks and almost nothing on the rest.
Second compressor: slow, optical, constant
An optical compressor, or one set with a ratio between 2:1 and 3:1, attack around 10 ms and a slow or program-dependent release. It should move little but all the time: 2–3 dB of constant reduction. This is what makes the vocal steady and “upfront” without squashing it.
De-esser after compression
Compression brings up sibilance, so the de-esser goes after it. Find the frequency of the “s” sounds between 5 and 8 kHz (between 4 and 10 kHz on some voices) and aim for 2–4 dB of reduction on the sibilants only. Listen in context: an over-de-essed vocal sounds toothless.
Colour EQ, after dynamics
Now you can add: a little presence between 2 and 5 kHz, soft air above 10 kHz, warmth between 150 and 250 Hz if the body is missing. Small moves, 1 to 3 dB. If you have to boost a lot, the problem is earlier, in the recording or the subtractive cut.
Parallel compression, if you need density
Send the vocal to a bus with an aggressive compressor (10 dB of reduction or more, fast attack) and blend it under the main vocal, usually between 10 and 20%. It adds body and presence without taking dynamics away. You do not always need it: on ballads it is often too much.
Automation: the last dB is yours
No compressor listens to the lyrics. After the chain, automate the level of the words that get lost and the tails that drop. They are often one-dB moves, but they make the difference between a compressed vocal and a vocal that speaks to the listener.
Compare at matched level and at low volume
Bypass the whole chain while setting the output so the perceived level is the same: only then can you tell whether you improved the vocal or just made it louder. Then listen to the mix at a very low volume. If you understand every word without effort, the compression works.
The starting values in a table
| Stage | Ratio | Attack | Release | Reduction |
|---|---|---|---|---|
| First compressor, fast | 4:1 – 8:1 | under 1 to 5 ms | 50 – 100 ms | 3 – 6 dB on the peaks |
| Second compressor, optical | 2:1 – 3:1 | about 10 ms | slow or program-dependent | 2 – 3 dB constant |
| De-esser | — | fast | fast | 2 – 4 dB on sibilants |
| Parallel bus | 4:1 or more | fast | medium | 10 dB or more, blended low |
You can also swap the first two stages, optical first and fast second: it is the order I often use in pop and R&B, where I want the slow compressor to shape the vocal and the fast one to catch only the last peaks. Both work: what matters is that each one does little.
These are starting points, not recipes. A soul vocal with big dynamic jumps asks for a firmer first stage; an already steady rap delivery sometimes does not need the second.
Two compressors doing a little sound better than one doing a lot. And the last dB always comes from automation.
Rap and trap: what changes
In rap the vocal is the rhythm of the track: it has to sit at the same level over the beat all the time. Here I push both stages harder and almost always use the parallel bus. Ad-libs get compressed even more and kept lower and narrower than the lead. If you use Auto-Tune as an effect, I cover it in detail in the article on mixing rap and trap vocals.
Signs you are over-compressing
- Breaths become as loud as the words.
- Sibilance grows even with the de-esser on.
- Mouth noises appear that were not there before.
- The vocal feels glued to a fixed level and loses intention: the intimate verses sound as loud as the chorus.
If it happens, go back one stage and lower the threshold, or let clip gain do more of the work.
The vocal is the centre of the track. Mix and master from $110, tuning included: timing, dynamics and effects shaped around your voice.
Start a project