Posts Tagged ‘ De-essing ’

Everything you wanted to know about de-essing but were too afraid to ask

Simply, de-essing is a process for reducing the level of sibilance in a vocal recording.

What is sibilance?

Sibilance is characterised by ‘sss’ and ‘ts’ sounds (and, to a lesser extent, ‘t’ and ‘k’ and ‘z’ sounds) in the English language. Unlike vowels, sibilant sounds have a relatively low (in volume) pitched component and a high (in volume) unpitched noise component. The unhitched noise is also focussed strongly in the higher register (unlike ‘shh’ sounds).

Why would you want to reduce it?

Sibilance is essential for intelligibility. That is, we need to hear it in order to understand the words delivered by the vocalist. Too much, however, can unbalance a mix. Some singers naturally deliver sibilant sounds loudly (this often applies to singers who aren’t classically trained). Sometimes EQ or compression can enhance the sibilance in a vocal recording (especially when the high frequencies are boosted).

When listening to the voice on its own it can be difficult to know if the sibilance is too strong. This is because our brain naturally compensates for the difference in volume between the vowels and the sibilance. In a mix, however, you’ll notice when the sibilance is too strong. You’ll raise the vocal level until the vowels are at the right level but the sibilance is too sharp and ‘sticking out’ of the mix, or your’ll reduce the vocal level until the sibilance sits well but the vowels disappear under the mix. Sometimes strong sibilance can excite the vocal reverb, making the reverb much more noticeable.

How does de-essing reduce sibilance?

De-essers are usually set up as simplified compressors with a bandpass or highpass filter in the sidechain.

Most de-essors do not have the full compliment of compression controls (attack, release, threshold, ratio, makeup). Instead, there is usually just a threshold (and sometimes a ratio control). The other controls are tuned to work with sibilance and the human voice. De-essers are among the most specialised studio tools – they don’t need a lot of controls or a wide range of operation.

The filtered sidechain changes the compressor’s behaviour so that it only reduces gain when there is sibilance in the audio. The filter is tuned quite high (usually above 5kHz) so that the compressor doesn’t respond to energy in the low or mids (where most of the vowel energy is). The compressor, however, applies gain reduction to the whole signal – not the filtered version. This means that when the sibilance is being reduced, the actual tone of the voice is not changed. It’s just made quieter.

Some de-essors use dynamic EQ instead of a compressor. They’re usually designed so that instead of reducing the gain of the whole signal, they only reduce gain to the high frequencies. Imagine a high shelf or parametric cut that only comes in when the voice is sibilant. This kinds of de-essers require more care when they’re being set up because they work by changing the tone of the voice (rather than just the level). If not configured well, they can make the vocalist sound like s/he has a lisp.

How do I set up a de-esser?

I usually wait until the mix is almost completely finished before applying a de-esser. I’ll make sure the level of the vowels in the vocal are balanced well against the other elements of the mix. I’ll then use the de-esser just enough to bring down the sibilance to an acceptable level. Usually, I aim for the lead vocal sibilance to be at a similar level to the hi-hats, snare or other prominent high-frequency sound in the mix.

I almost almost insert the de-esser after EQ and compression (but before any time-based effects such as delay or reverb, of course). This is because I use the de-esser to slightly modify a sound that I’m already happy with. Applying compression after a de-esser can actually counteract the de-essing, as the rull-range compressor can bring the sibilance level back up.

What else is a de-esser useful for?

De-essers can be very useful for backing vocals. There are some situations where backing vocals (especially stacked backing vocals) are a little messy. Most of the time it’s fine, but it’s most pronounced in the sibilance. A de-esser will bring the sibilance right down, making the backing vocals sound less messy. Use this way, a de-esser can be applied much more heavily – the intelligibility and articulation is carried by the lead vocals. Just watch out for any sections where the background vocals are exposed – heavy de-essing will make them sound weird without the lead vocal.

De-essing can sometimes be useful on drum kits – particularly on overheads when balance of the kit is right but the crash cymbals are too loud. A de-esser can sometimes be effective in reducing the level of the crash cymbals while still retaining the sense of room and space. Again – subtlety is the key here. Too much de-essing will suck the air out and make the drums sound unnatural.



Vocal processing

My usual vocal processing chain consists of several stages: Gate, EQ, Compression, De-Essing, and reverb (as a send).


This is first in the chain so the gate has the full dynamic range of the original audio. The more natural the dynamic range available to the gate, but easier it is to set the threshold and timing for a natural sound.


Generally I prefer to use EQ before compression. This is so I can get the tone I want for the mix before I adjust the dynamic range. It also makes it easy to highpass the audio so the compressor doesn’t respond to low-frequency audio (such as rumble) that isn’t going to make it to the mix anyway. I’ve written more about the order of EQ and compression here.


I choose to apply compression to the final tone of the sound, rather than adjust the tone afterwards. This helps the compressor react smoothly and naturally to the sound we hear, rather than responding to sound that is going to have its frequency balance changed afterwards.


I don’t often use saturation of vocals. When I do though, it’s just after compression,  and I use it similar to a limiter  – to catch the few peaks that are too loud even after compression. Usually I set it up so that loud sustained notes are saturated, making them sound loud without overpowering the mix, but most notes are left clean (not saturated).


This is interesting. I’ve found that I get the best results by applying the de-esser after EQ and Compression. I find that the way I use EQ tends to enhance sibilance (tonal tilt toward high frequencies and high-ratio compression). Using a de-esser earlier in the chain sometimes means that the later EQ and compression counteract the effect of the de-esser. This forces me to apply more de-essing, which ends up sounding (more) unnatural (at extremes, it can “pump” a bit – but not in a good way!). By de-essing after EQ and compression, only a silght amount of de-essing is needed.


Reverb can be quite sensitive – responding to both tone and dynamics. In the kind of dense productions I usually do, it’s best to feed the reverb as consistent a signal as possible, to avoid widely-warying levels of ambience or a build up of mud. The purpose of reverb here is to add ambience and air to the vocal sound – not to be heard as an effect separate to the vocal. To that end it’s important that the reverb responds to what we hear (similar to the compressor). I’ll often lowpass the reverb to keep it sounding lush and avoid it “catching” any sibilance.