Sibilance

Sometimes a vocal can sound very irritating—listen to 10 s of this (start at 4:01):

That “S” sound at the end of a word is called “sibilance”. It occurs typically in the 4 to 8 kHz range; that same range is needed to get clarity and presence in a vocal. Too much, almost like symbols, very quickly becomes very irritating. However, too little, and a person battle to hear what is being said.

From the clip: This is the 'wy' of 'wys' in 'Wil ek vir jou wys wat leer...': the frequencies shown at the bottom correspond to the yellow cursor at the top.

This is the 's' of 'wys' in 'Wil ek vir jou wys wat leer...'—note the portion that is significantly louder. The 5 kHz peaks at –50 dB, which was previously –63 dB—that’s a 13 dB difference, which represents more than four times the volume (6 dB is a doubling).

The process of removing sibiliance is called “de-essing” or “desibilising” (https://en.wikipedia.org/wiki/De-essing).

The first step in the process of improving sound is normally recording better quality sound at the sound, thus placing the microphone further away (which Marthinus already does), and aiming the microphone higher or lower (away from the mouth), as you can see in, for e.g. https://www.soundonsound.com/techniques/managing-sibilance or https://urm.academy/death-to-sibilance/.

If one already has the sound, you can process it via an EQ. The challenge with this is that taking away too much of the “s” sound results in losing clarity and presence. Thus one wants to only take away the problematic portions when they are excessive, but leave them the rest of the time, i.e. compress that which annoys, but only when necessary.

Reaper’s built-in 'De-esser'

Reaper’s “De-esser” is a single-band dynamic EQ. In the DiGiCo S31 we have this same functionality available via a dynamic EQ:

DiGiCo's S31 dynamic EQ: Here, the green band has the dynamic EQ turned on; here, around 5 kHz, the volume is currently being lowered. For the purpose of de-essing, the mode (top right) must be set to 'Over' so that when the level at 5 kHz exceeds the threshold, then the level of the 5 kHz band is reduced.

General guidelines

To find the most appropriate centre frequency: while sweeping, boost the frequency so that it sound worse, because it’s easier to hear increased irritation than a reduction of irritation. In the S31-world one can do this in a live environment by routing the same input to an unused channel, and then mute and solo that channel, and listen for the most problematic frequency in the headphones. At Brooklyn Campus specifically, there is a channel named “Test” that you can use for this purpose.

The dynamic EQ threshold needs to be high enough to not influence other sounds too much, but low enough to result in a significant improvement in the sibilance.

The compressor must typically be quite fast, i.e. short times, to sound as natural as possible.