I recently did an overview identifying what is multiband compression and talked a bit about the various applications for it in mixing. One of the many applications of this frequency specific compressor is to use it as a de-esser – a tool primarily used to attenuate vocal sibilance. Rather than getting a specific de-essing plugin, let’s talk how to make a stock plugin which any DAW has in a multiband compressor a de-esser.
How to Make a Multiband Compressor a De-Esser
As I just alluded to, most DAWs these days don’t have a de-esser as a stock plugin. Virtually any DAW you would use will come with a multiband compressor, so we can turn this into a de-esser with some specific settings.
A de-esser is essentially just a multiband compressor anyway; the only difference is some de-essers like my favorite FabFilter Pro-DS are specifically designed to pick out and treat sibilance (those harsh S and T sounds on vocals).
In this example for making a multiband compressor a de-esser I’m using FabFilter’s Pro-MB, but the same settings will work with your DAW’s stock multiband compressor/your multiband compressor of choice.
Here are the settings for turning a multiband compressor into a de-esser (we’ll next break down each specific setting):
Multiband Compressor De-Esser Settings
Frequency – The peak frequency of sibilance is around 7-8k, but it can go as low as 5k and as high as 12k. As such, set your frequency band accordingly as I show above. Note that just like with EQing vocals, the exact target range tends to be a little higher with female vocalists and a little lower with male vocalists, not to mention the vocalist’s range in general.
Solo around 7-8k and sweep around to hear the peaks and most offending parts of the sibilance to nail the top end.
Threshold/Range – As compared to normal compression, when we’re targeting sibilance in particular, we generally can get away with more gain reduction while keeping the vocal sounding natural.
Part of the reason we can use multiband compressors as de-essers like we’re detailing here is that the biggest surges in that 7-8k region occur on sibilant moments. This makes setting the threshold very important.
As such, find the most offending sibilant part of your vocal, probably the worst S sound, and aim for 8-10dB of gain reduction max. This will smooth out the rest of the sibilance in your vocal at the less extreme parts.
Remember the goal is just to smooth the sibilance out, not remove it altogether. This is because one, a certain amount of sibilance is normal on these consonants, and two, over compressing at this frequency will make the vocalist sound like they have a lisp.
Ratio – The ratio is key here. Set a relatively aggressive compressor ratio of 8:1 and a hard knee of -6dB to ensure that when the sibilance pushes that frequency range over the threshold even if only slightly, it gets significantly attenuated.
Attack/Release– Set the attack and delay to be fast to instant. Set alongside the threshold, range, and ratio, this will work well for de-essing. The transients of vocals are lower than that 7-8k range, so no need to worry about taking away the bite from your vocal.
Lookahead – Add at least 1ms of lookahead time to get a more transparent and effective de-essing from your multiband compressor without sacrificing much to any mixing CPU.
Once you tweak these settings to perfectly fit your vocal, you can save that as a preset to know that you’ll have a de-esser in your multiband compressor whenever you need it to keep that sibilance in check.
Don’t limit your use of de-essing to vocals, however, as this tool can be applied to any elements which you find are leading to a harsh mix.
I like to use it on guitars, synths, tambourines, cymbals – anything which might be a bit too bright. You may need to tweak your settings, but the above settings provide a solid framework for smoothing out any tracks which need it.