What is a vocal chain?
A vocal chain is the list of plug-ins on a vocal track, in the order the sound passes through them. In a DAW they're the insert slots on the vocal's channel: the sound goes into the top slot, comes out into the next one down, and carries on to the fader. Each plug-in hears whatever the one above it did, so two chains with the same plug-ins in a different order can sound quite different.
The stages on this page each do one job. The high-pass filter and the first EQ take away sound you never wanted, like rumble or a boxy room tone. The compressor and the de-esser control level, so every word can be heard over the music and no s stings. Saturation adds density, and the last EQ sets how bright and forward the voice sounds. Reverb and delay put the voice in a space, and they usually sit on separate return tracks fed by a send, outside the chain itself.
None of this is a rule. A vocal recorded in a treated room on a good microphone might need only a compressor and a little EQ, while a demo recorded on a laptop in a bedroom might need every stage here. The chain in the lab is a sensible starting order for a lead vocal in pop, R&B, rock or electronic music, and the sections below say what each stage is for so you can tell when to leave one out.
Reading the lab
The row of six blocks under the sound picker is the chain, in signal order from left to right. Click a block to switch that stage on or off. Each block shows what its stage is doing to the loop right now, measured from the sound itself, and the knobs for each stage sit in their own group below the screens.
The big screen shows the whole loop. The pale shape is the vocal going in and the pink shape is what comes out of the chain, drawn at the same loudness, so any difference in shape is the chain's doing. Where the pink shape is flatter than the pale one, the loud words have been brought closer to the quiet ones. The amber lane underneath is the compressor's gain reduction.
The EQ screen adds up what the filters are doing to each frequency. The pink line is the high-pass, the cut and the tone stage together, and the dashed teal line is the de-esser, which only moves while an s is coming through; watch it dip in time with the esses while the loop plays. The meters show the level going in, the gain reduction of the compressor and of the de-esser, and the level coming out.
Press A/B to switch the whole chain off. The lab sets the original to the same loudness as the processed version, because a vocal that's simply louder sounds better at first and would win every comparison. The Right now panel explains what your current settings are doing, using numbers measured from the loop.
The two vocals are public domain recordings, but the lab is most useful with your own voice. Press Record and sing for eight seconds, or upload a file, and it loops through the chain in place of ours. Nothing leaves your device.
Before the first plug-in
A chain works best on a vocal that's already in good shape, and a few minutes of preparation saves a lot of plug-in work. Record at a sensible level first. At 24-bit, peaks around −10 to −6 dBFS leave plenty of headroom for a loud chorus and keep the noise floor far below the voice; the gain staging lesson explains why recording hotter gains you nothing.
Then even out the big jumps by hand. If one phrase is 6 dB quieter than the rest because the singer stepped back from the microphone, turn that region up with clip gain (every major DAW lets you change the level of a single clip or region) before it reaches the compressor. A compressor set hard enough to fix a 6 dB jump would squash every other word as well.
Cut out loud breaths, clicks and bumps between phrases at this stage too. Leave the quiet breaths in, since a vocal with every breath removed sounds unnatural, but anything that would make the compressor or the de-esser react is better gone.
The vocal chain order
This is the order the lab uses, and the one I'd start any lead vocal with. The settings are starting points for a typical close-miked pop vocal, and the sections after the table go through each stage.
| Stage | What it fixes | Start here | Too much sounds like |
|---|---|---|---|
| 1. High-pass filter | Rumble, thumps, close-mic boom | 80–120 Hz, 12 dB/oct | A thin, small voice |
| 2. Subtractive EQ | Boxy or muddy ranges | −2 to −4 dB, 200–800 Hz | A hollow voice |
| 3. Compressor | Words that jump out or vanish | 3:1–4:1, 3–6 dB GR | Loud breaths, a flat performance |
| 4. De-esser | Sharp s, sh and t sounds | 5–9 kHz, 3–6 dB off the worst | A lisp, s turning into th |
| 5. Saturation | A thin or distant voice | Light drive | Fuzz on the loud notes |
| 6. Tone EQ | Not enough presence or air | +2 to +4 dB at 3 kHz, shelf above 10 kHz | Harsh consonants, hiss |
| Sends | No sense of space | Reverb with 20–80 ms pre-delay | Blurred words |
1. High-pass filter
A high-pass filter removes everything below its cutoff. A voice has very little useful sound under about 80 Hz, but a vocal track picks up plenty down there: rumble from traffic or air conditioning, the singer knocking the stand, and pops from p and b sounds. Directional microphones also boost the bass when the singer is close to them, which is called the proximity effect, and that extra low end makes a vocal boomy.
Start at 80 to 120 Hz with a 12 dB per octave slope, which is what the lab uses. Then raise the frequency while the vocal plays with the rest of the song, stop as soon as the voice sounds thinner, and back it off a little from there. A deep male voice might want 60 or 70 Hz. A light, breathy voice can often go up to 150 Hz without losing anything.
In the lab, the high-pass at 100 Hz takes out under 1% of the recording's energy, because this voice had very little below 100 Hz to begin with. Record your own voice into the lab with the microphone a few centimetres from your mouth and you'll see the high-pass do a lot more.
The high-pass and low-pass filters lesson covers cutoffs and slopes, and low cut vs low shelf explains the gentler option.
2. Subtractive EQ
Next, take out anything that sounds wrong. Every voice and room is different, but the problems collect in familiar places. Mud usually sits between about 200 and 300 Hz. A boxy, cardboard sound between 300 and 800 Hz is common when a vocal is recorded in a small untreated room, and a nasal honk tends to show up around 1 kHz.
To find the problem, set an EQ bell to a narrow width, boost it by 6 to 10 dB and sweep it slowly through that range until the bad sound jumps out. Then turn the boost into a cut of 2 to 4 dB and widen the Q until it sounds natural. A big boost exaggerates everything it passes over, so cut less than the sweep makes you want to.
The lab's cut is about an octave wide. On the default vocal, the octave around 500 Hz holds more than a third of the voice's energy, so a 3 dB cut there changes the tone more than the number suggests.
Doing this before the compressor is the reason it's second in the chain. A compressor reacts to whatever is loudest, and if that's a boxy 500 Hz resonance, it turns the whole voice down every time the singer hits a note that sets it off.
Subtractive EQ covers the technique on its own, and EQ Q and bandwidth explains the width control.
3. Compressor
The compressor's job on a vocal is to keep every word audible. Singers get louder on high notes and quieter at the ends of phrases, and over a busy track the quiet words disappear while the loud ones jump out. A compressor turns the loud moments down so the gap between them shrinks, and then the whole vocal can sit at one fader level. The compression lesson explains every control in detail.
Start at 3:1 or 4:1 with an attack of 5 to 15 ms and a release of 50 to 150 ms. Then lower the threshold until the loudest words show 3 to 6 dB of gain reduction. The lab fixes the attack at 8 ms and the release at 120 ms, so you can concentrate on the threshold and the ratio, and it adds makeup gain automatically.
With the threshold at −16 dB and a 4:1 ratio, the default vocal's loudest words come down by up to 5.3 dB. The compressor is working for about a third of the loop. To show what that does to the performance, the lab's Loud to quiet figure measures how far apart the loud and quiet moments of the sung parts are, in 50 ms slices, and it drops from 12.8 dB to 10.1 dB. Those 2.7 dB come off the loudest words, the ones that would otherwise poke out of the mix.
You can also split this job between two compressors in a row. The first is fast and only catches the sharpest peaks; a FET design such as the Universal Audio 1176 is a common choice. Then a slower optical one, such as the Teletronix LA-2A, smooths out the performance. Each takes off a few decibels, which tends to sound more natural than one compressor doing all of it. Parallel vs serial compression compares this with blending in a squashed copy of the vocal.
4. De-esser
Sibilance, the hiss of s, sh and t sounds, usually sits between 5 and 9 kHz. A de-esser is a compressor that listens only to that range and turns it down while an s is coming through, so the rest of the voice keeps its top end. The de-essing lesson covers it in depth.
It goes after the compressor because compression makes sibilance worse. The compressor turns the loud vowels down, then makeup gain lifts everything back up, including the esses, which are quieter than the vowels and so set off less gain reduction of their own. In the lab, the band above 6 kHz peaks at −8.1 dBFS before the compressor and at −3.9 dBFS after it.
Set the threshold so only the esses cross it, and aim for 3 to 6 dB off the worst ones. At −18 dB, the lab's de-esser catches 5 esses in the loop and turns the deepest down by 6.6 dB, and it works for 7% of the time. If yours is working for much more of the song than that, it's taking the air out of words that had no sibilance in them.
One exception is a singer whose esses are so loud that they make the compressor jump. Then put a gentle de-esser before the compressor as well, and keep the main one after it.
5. Saturation
Saturation is gentle distortion. It adds harmonics above each note and rounds off the peaks, which makes a voice sound denser and a little closer. Tape and valve emulations are the usual choice, and the saturation lesson explains how the different curves sound.
Putting it after the compressor means the saturator sees a more even level, so the loud notes don't distort much more than the quiet ones. In the lab, 8 dB of drive into a soft curve adds distortion that sits about 35 dB below the voice. You hear that as thickness more than fuzz, and it brings the peaks down another 0.9 dB against the average level.
Turn the drive past 12 dB and you start to hear the distortion itself. That suits some rock and hip-hop vocals. If you want the density without the fuzz, blend the saturator in with its mix control, the way parallel saturation describes.
6. Tone EQ
The last EQ is for boosts. Presence lives between about 2 and 5 kHz, where the consonants that make words easy to follow sit, and a couple of decibels there pulls a vocal forward in the mix. Air is the breath and shine above 10 kHz, usually added with a high shelf.
Boosting after the compressor stops it reacting to the extra top end, and boosting after the de-esser keeps the de-esser's threshold where you set it. In the lab, +3 dB at 3 kHz and +4 dB of air raise the share of the vocal's energy above 5 kHz from about 6% to 7%. The default recording has almost nothing above 11 kHz, so here the air shelf mostly brightens the range between 6 and 11 kHz; on a recording with more top end it reaches further up.
Listen to the esses again after this stage, because the air boost lifts them a little too. If they've started to sting, lower the de-esser's threshold.
Reverb and delay on sends
Reverb and delay come after the chain, usually on their own return tracks. A send copies the processed vocal to the return, the effect runs there at 100% wet, and the send level sets how much of it you hear. The dry vocal stays untouched, you can EQ the reverb on its own, and the backing vocals can share the same space. The insert vs send lesson compares the two ways of routing.
A few settings help a vocal stay clear inside its reverb. Use 20 to 80 ms of pre-delay so the start of each word arrives before the reverb does. High-pass the return at around 200 to 300 Hz and take a little off the top, so the reverb doesn't add mud or another layer of sibilance. For delay, a quarter-note or eighth-note echo synced to the tempo is a good place to start, which is 500 ms or 250 ms at 120 BPM, and a ducked delay keeps the echoes down while the singer is singing. The lab stops before the sends, and the reverb lesson has a lab of its own.
Does the order of a vocal chain matter?
Yes, because each plug-in reacts to what the one before it did. The high-pass and the two EQ stages would sound the same in any order among themselves, but the compressor, the de-esser and the saturator all behave differently depending on what reaches them.
EQ cuts go before the compressor so it doesn't react to mud and resonances, and EQ boosts go after it so it doesn't fight them. If you boosted 3 kHz before the compressor, it would turn the vocal down harder on the bright words and undo part of your boost. EQ before or after compression walks through both cases.
A saturator after the compressor sees a steadier level, so it distorts every word by a similar amount. Put it first and the loud notes distort far more than the quiet ones, which can be the sound you want on an aggressive vocal.
Plenty of engineers move things around. Some put the saturator before the compressor, and some use one de-esser early and another late. Both work. The order in the lab is a reliable place to start, and effect order covers the reasoning for other instruments too.
How to build a vocal chain, step by step
- Work with the music playingA vocal that sounds great on its own can disappear under the track. Solo it to hunt for a problem, then make the decision with everything playing.
- Fix the level by handClip gain any phrase that's much louder or quieter than the rest, and cut out loud breaths, clicks and bumps.
- High-pass, then cutRaise the high-pass until the voice thins and back it off. Sweep for mud or boxiness and take out 2 to 4 dB.
- CompressStart at 3:1 or 4:1 and lower the threshold until the loudest words show 3 to 6 dB of gain reduction.
- De-essSet the threshold so only the esses cross it, with 3 to 6 dB off the worst of them.
- Add colour and toneA little saturation, then presence around 3 kHz and air above 10 kHz until the voice sits in front of the music.
- Send it to reverb and delay, then check each stageBypass every plug-in in turn with its output set so on and off are equally loud. Anything you can't hear helping comes out.
Common vocal chain mistakes
Judging plug-ins by how loud they make the vocal
Most stages in a chain change the level, and a vocal that's a decibel louder usually sounds better for a moment. Compare every plug-in at matched loudness, either with the A/B in the lab or by setting each plug-in's output so that switching it off doesn't change the volume.
Filtering out the body of the voice
A high-pass at 200 Hz can sound clean on a soloed vocal and thin as soon as the music comes in. Set the filter with the track playing and stop at the first sign of thinning.
Too much compression
Push the compressor hard and every quiet sound comes up with the quiet words: breaths, lip noise, the room. In the lab's last step it takes off up to 27 dB and the loud and quiet moments end up 3.6 dB apart, which is about as flat as a voice gets. If you need that much control, use two compressors, or fix the level by hand first.
Keeping stages you can't hear
Each stage should fix something you can point to. If bypassing one at matched loudness changes nothing you care about, take it out, even if it's a plug-in you paid for. Some plug-ins also add latency, which you'll notice if the singer monitors through the chain while recording.
Copying someone else's chain
A chain is a set of fixes for one recording. A preset made for another singer in another room will have its high-pass, cut and thresholds in the wrong places for yours. Keep its order if you like it, and set every stage again by ear.
Train your ears
The quiz under the lab plays the vocal through a full chain at matched loudness, sometimes with one stage switched off, and asks which one is missing. The compressor shows up as words that jump out again, the de-esser as esses that sting, and the tone stage as a voice that sits further back. Eight rounds take a few minutes, and your best score is saved in your browser. Try it again with your own recording once you've done a round with ours.
Questions people ask
What order should plug-ins go in on a vocal?
A good default is high-pass filter, subtractive EQ, compressor, de-esser, saturation, then a final EQ for presence and air, with reverb and delay on sends. Cuts go before the compressor so it doesn't react to mud; boosts go after it so it doesn't fight them.
Should the de-esser go before or after the compressor?
After, in most chains. Compression and makeup gain bring the esses up, so a de-esser after the compressor deals with the sibilance you actually hear. If loud esses are also making the compressor jump, add a gentle de-esser before it as well.
Should reverb go on the vocal track or on a send?
On a send. The vocal stays dry and clear on its own channel, the reverb return can be filtered on its own, and backing vocals can share the same reverb. A reverb in an insert slot with its mix control turned down works too, if that one vocal is the only thing that needs it.
Can I use a vocal chain preset?
As a starting point, yes. A preset was set up for a different voice, microphone and room, so expect to change the high-pass frequency, the EQ cut and the compressor and de-esser thresholds for your recording. The order is the part of a preset most worth keeping.
Do I need paid plug-ins for a good vocal chain?
No. Every stage on this page can be done with the plug-ins that come with a DAW; in Logic Pro, for example, Channel EQ, Compressor and DeEsser 2 cover the first four stages. Setting the stages well makes far more difference than which brand they are.
How loud should a vocal be before the chain?
When recording at 24-bit, aim for peaks around −10 to −6 dBFS. That leaves headroom for loud notes and keeps the noise floor far below the voice. Turning the input up further gains nothing and risks clipping the converter on the loudest words.