How to compress vocals in Ableton Live
To compress a vocal, use two gentle stages instead of one hard one. Catch the loudest words with a fast attack and a 4:1 to 8:1 ratio, then level the phrases with a 2:1 to 4:1 ratio, a 10 to 30 ms attack and auto release, aiming for 3 to 6 dB of gain reduction.
| Stage | Detector | Ratio | Attack | Release | Gain reduction | Threshold, peaks at −6 dB |
|---|---|---|---|---|---|---|
| Catch the peaks | Peak | 4:1 to 8:1 | 1 to 5 ms | 40 to 80 ms | 2 to 4 dB | −10 dB |
| Level the phrases | RMS | 2:1 to 4:1 | 10 to 30 ms | Auto | 3 to 6 dB | −20 dB |
| De-ess (optional) | Peak | 4:1 | 0.5 ms | 60 ms | 3 to 6 dB | −20 dB |
Vocals use 3 stages on this page: catch the peaks, level the phrases and de-ess.
2 stages are part of the starting point, and 1 more stage waits until you hear a problem.
Across them the ratio runs from 2:1 to 8:1 and the gain reduction targets from 2 to 6 dB. Set every threshold last, by watching the gain reduction meter, not by copying a number.
The stages, in order
Catch the peaks, 4:1 to 8:1
A few loud words jump out of the mix. Grabs only the loudest words, so the next stage gets a steadier signal.
Start with the Peak detector, a ratio of 4:1 to 8:1, an attack of 1 to 5 ms and a release of 40 to 80 ms. The fast attack catches each peak before it gets past. Set the release so the meter falls back to zero before the next note or hit, or the next one starts already turned down and loses its punch.
Lower the threshold until the meter shows 2 to 4 dB of gain reduction on the loudest parts. If a vocal peaks near −6 dB, that is a threshold of about −10 dB.
Level the phrases, 2:1 to 4:1
Quiet words get lost and held notes push forward. Evens out whole phrases so the vocal sits at one level.
Start with the RMS detector, a ratio of 2:1 to 4:1, an attack of 10 to 30 ms and auto release. The slow attack lets the front of each note or hit through before the compressor reacts, which keeps the punch. Auto release follows the material, letting go quickly after short peaks and slowly after long ones.
Lower the threshold until the meter shows 3 to 6 dB of gain reduction on the loudest parts. If a vocal peaks near −6 dB, that is a threshold of about −20 dB. The RMS detector reads the average level, which sits a few dB below the peaks, so its threshold sits lower.
A soft knee of 6 dB makes the compressor ease in, so the change is harder to hear.
De-ess, 4:1
S and T sounds hurt on earbuds. Turns down only the sharp S and T sounds, because its sidechain EQ near 7 kHz means it only hears that band.
Start with the Peak detector, a ratio of 4:1, an attack of 0.5 ms and a release of 60 ms. The fast attack catches each peak before it gets past. Set the release so the meter falls back to zero before the next note or hit, or the next one starts already turned down and loses its punch.
Lower the threshold until the meter shows 3 to 6 dB of gain reduction on the loudest parts. If a vocal peaks near −6 dB, that is a threshold of about −20 dB.
This stage stays out of the starting point. Add it only when you can hear the problem in a vocal.
The settings in Live 12
CompressorLite, Intro, Standard, Suite
- Model
- RMS
- Threshold
- -20.0 dB
- Ratio
- 3.00 : 1
- Attack
- 10.0 ms
- Auto Release On/Off
- On
- Knee
- 6.0 dB
- Makeup
- On
This is the leveling stage. Model, Ratio, Attack, Auto Release and Knee match the leveller in the kit's Vocal Chain. For the peak stage, put a second Compressor before it with Model on Peak, Ratio at 6.00 : 1, Attack at 3.00 ms and Release at 60.0 ms.
Glue CompressorStandard, Suite
- Threshold
- -15.0 dB
- Ratio
- 2
- Attack
- 10
- Release
- A
- Makeup
- 3.00 dB
Glue Compressor stands in for the leveling stage, after a Compressor set up as the peak stage. Its lowest ratio is 2:1, and Release on A follows the phrasing for you.
In short: catch the peaks at 4:1 to 8:1 with 2 to 4 dB of gain reduction and level the phrases at 2:1 to 4:1 with 3 to 6 dB of gain reduction. Then switch the compressor off and on at the same loudness, and keep it only if the part sounds steadier, not just louder.
Even out the level before you compress
A compressor works best on a vocal that is already close to even. If the second verse was sung a long way from the mic, or one line is far louder than the rest, the compressor has to work very hard on some phrases and hardly at all on others. You hear that as a vocal that sounds fine in one place and squashed in the next.
So do the big moves by hand first. In Live, split the clip at the phrase that jumps out and turn that piece down with clip gain, or draw clip gain automation across the song. You are not trying to make every word the same level, only to get each phrase into the same rough area. Then the compressor only has to smooth out the small jumps, which it does well.
This is also the time to cut what you don't want the compressor to hear. Rumble, pops and a muddy low end all push the compressor around, so put your EQ cuts first in the chain. A compressor that is reacting to low end you were going to cut anyway will turn the words down at the wrong moments.
Two stages instead of one
One compressor doing all the work has to be set hard, and a hard compressor on a vocal sounds pinched. Two gentle stages, each doing a small job, sound far more natural.
The first stage catches the peaks. It uses the Peak detector, a fast attack and a ratio high enough to hold the loudest words, and it should only move on those words. Most of the time its meter sits still. When a word jumps out, it moves a few dB and then lets go fast, before the next word starts.
The second stage levels the phrases. It uses the RMS detector, which listens to the average level instead of every peak, with a gentle ratio, a slower attack and auto release. This is the stage that makes the vocal sit at one level against the beat. Because the first stage has already caught the spikes, this one can stay gentle and still hold the whole vocal in place. The free kit's Vocal Chain levels a vocal the same way, with an RMS Compressor at 3:1 and a 10 ms attack under its Squash knob.
Sibilance is a third, optional job. Compression brings up the quiet parts of a vocal, and the S and T sounds come up with them. If they start to hurt, add a de-esser. In Live that is a Compressor with its sidechain EQ focused near 7 kHz, so it only reacts to the sharp sounds. The kit's Vocal Chain does this with its De-ess knob, at 4:1 with a 0.5 ms attack. Leave it out until you can hear the problem, because a de-esser working on a vocal that doesn't need one makes it lisp.
Set each threshold by ear and by the meter. Play the loudest part of the song and lower the threshold until the meter shows the gain reduction you want. The settings on this page are a place to start, and your ears decide the rest.
Rap, pop and backing vocals
A rap vocal carries the rhythm, and the words have to land clearly over a busy beat. It usually takes a little more compression than a sung vocal, and an attack near 10 ms on the leveling stage keeps the consonants tight. Watch the breaths, because the harder you level a rap vocal, the louder the gaps between lines get.
Pop lead vocals sit close and bright, and modern pop often levels them hard so every word is right in front. Get there with the two stages working together rather than by pushing one. If the vocal still jumps out in the chorus, a third pass of clip gain often works better than more compression.
Backing vocals should sit behind the lead, so they can take more compression than the lead does. A steady, flat backing part blends into a soft layer and stays out of the way. Group them on one bus and compress the bus, so they move together like one voice.
Mistakes that make vocals sound squashed
Too much gain reduction is the main one. If the leveling stage moves much more than 6 dB on the loudest phrases, the vocal starts to sound flat and lifeless, and the breaths and mouth noise come up between the words. Raise the threshold until the meter moves less.
An attack that is too fast on the leveling stage is the next. It grabs the front of every word and makes the vocal sound dull and far away. The fast attack belongs on the peak stage, which only moves on a few words. Keep the leveling stage at 10 ms or slower.
Judging louder as better is easy to fall into. Makeup gain makes the compressed vocal louder, and louder sounds better at first. Match the level before you compare with the compressor off.
Compressing in solo is the last. A vocal that sounds squashed on its own can sit perfectly in the mix, and one that sounds open in solo can disappear behind the guitars. Make every final call with the whole track playing.
Questions people ask
What does vocal compression do?
Vocal compression turns the loudest words down so the quiet ones can sit closer to them. The vocal then holds one steady level against the beat, and you can hear every word without riding the fader. On this page that takes two stages, a fast peak catcher and a slower leveller.
How much should you compress a vocal?
Aim for 3 to 6 dB of gain reduction on the loudest phrases from the leveling stage, plus 2 to 4 dB from a peak catcher on the words that jump out. If the meter moves much more than that, the vocal starts to sound flat and the breaths come up.
What are the best vocal compression settings?
A good start is two stages. First a Peak compressor at 4:1 to 8:1 with a 1 to 5 ms attack and a 40 to 80 ms release, then an RMS compressor at 2:1 to 4:1 with a 10 to 30 ms attack and auto release. Set each threshold by watching the gain reduction meter.
How do you set up a compressor for vocals?
Put Compressor after your EQ cuts on the vocal track. Set Model to RMS, Ratio to 3.00 : 1, Attack to 10.0 ms and turn on Auto Release, then lower Threshold until the meter shows 3 to 6 dB on the loudest phrases. Turn on Makeup and compare with the compressor off at the same loudness.