Pristine vocals deserve top notch processing; the quest for creating the best possible de-esser



De-essers can be an evil necessity. Vocal recordings may be too sibilant requiring de-essing (or excess sibilance removal), but most de-essers come with very clearly audible drawbacks as well. After de-essing, vocals may sound muffled, the ‘s’ may sound more like an ‘f’, or even worse, the operation of a de-esser manifests itself as a clearly-audible time-varying filter.

Yet another de-esser?
ToneBoosters already provides two de-essers; the TrackEssentials de-esser with zero latency, and an integrated de-esser in TB Evoke. Both plugins have their own merit, but we thought we should be able to do better. New TB technology, such as ultra-transparent gain processing and tonal decomposition (as featured in the recently overhauled TB BusCompressor), the matched-filter technology from TB Evoke, and a couple of new innovations together should, when combined properly, result in a much better de-esser.

De-essing like a compressor
You may see a very familiar input/output curve in the screenshot that looks like a compressor. In this case, the input/output curve does not relate to level, but to (excess) sibilance. Sibilance is a property of audio that is largely independent of level; signals sound sibilant if there is a relatively large amount of signal energy present in the sibilant range (typically 4-11 kHz) compared to the overall level. Nevertheless, TB Sibalance allows control of sibilance by means of a threshold, a ratio, a soft knee, and a range parameter; much like a compressor. Of course, a dry/wet control is included as well.

Algorithm fusion
TB Sibalance has three algorithms: (1) a broad-band de-esser, (2) a band-limited de-esser, and (3) a matched-filter de-esser. The latter will create a filter dynamically that only reduces sibilant frequencies while leaving everything else untouched. Contrast to many other de-essers, these algorithms can be fused on a continuous scale. Do you want 60% of de-essing using a band-limited de-esser, and the remaining 40% using matched filter technology? Just set the algorithm slider to 1.6 and it is all set. Moving the algorithm slider from 0, to 1, to a value of two seamlessly fuses the broad-band de-esser, the single-band de-esser, and the matched-filter de-esser.

Tonal component sensitivity
In many practical situations, vocal sibilance consists of noise-like signals that one would like to suppress. Tonal or voiced signals, on the other hand, are often better left untouched. Conventional de-essers cannot discriminate between noise-like and harmonic signals – they simply measure energies. With TB Sibalance, the relative contribution of tonal components to the measured sibilance can be adjusted so that the de-esser works much more accurately.

Mid only and high-quality modes
If a buss signal or full track needs de-essing, great results can be obtained by de-essing the mid channel only (this is typically where the vocals are), while leaving the side signals untouched. This mode of operation is available on TB Sibalance, as well as a control to engage a high-quality mode.

Signal level dependencies
It can be very desirable to reduce excess sibilance for relatively loud parts of a track, while leaving less loud elements untouched. The ‘level threshold’ control of TB Sibalance influences how sibilant low-level signals are. Basically, if the input signals approach the threshold set for level, the measured sibilance will gradually be reduced, and hence the amount of de-essing will become more subtle or even absent.

Processing of full mixes
Ideally, a sibilance tool is sufficiently flexible to also process full mixes, for example to catch excess esses in a mix, or simply to reduce the mix’s harshness. We’re currently experimenting with this, and this is why you’ll see controls to change the ‘voiced’ frequency range analysis; for mix processing, these can be set to cover almost the complete audible frequency range.

A public beta of TB Sibalance can be downloaded HERE (Windows x86 VST only!) in case you’d like to give it a spin.

More INFO: ToneBoosters | TB Sibalance

 

and I was saying

 

I am currently working on a "Streettape" for a homie of mine: you know, rapping over existing instrumentals, getting the vocals to sit right in the mix. Which is kind of challenging but good training…

Anyway, I had to de-ess his vocals and I tested a lot of De-Essers: C1 Sidechain Comp, bx Dyn EQ in De-esser Mode, R Deesser, DMG Essence and Fabfilter Pro-DS. Since DMG said, "I have the worlds most bestest Deesser" I was really hyped about it. Although it has a lot of visual feedback and a lot of controls I didn't manage to get it to work for me as I wanted it which was pretty frustrating.

So I ended up using Fabfilter Pro DS which was the best use for that scenario. It's really well designed and easy to control and has great sidechain monitoring - one for the threshold and one for the effected frequency. So as of yet, I'd say, Pro-DS is the best for and the one to beat. The one from Sonnox is probably the most advanced but let's see what can TB can come up with. Usually I only like their Tape stuff (Saturation and Reel Bus)…

 

Spectral editing/manipulation is an old idea.
There is Melda spectral dynamics, and a tutorial video on Youtube. Works but I find detailed work with Melda plugs PITA. Nice because based on a freuqncy compression.

And the Holy Grail of deesing: RX advanced. Paint by numbers. Marque the "S" press the button, save and repeat. Killer.
I didn't use de-esser for the last 3 years..
Cubendo user should take a look at : http://vstxmltranslator.blogspot.com/ which bypasses the shell/windows handling. Just copy anything from/to Cubendo to external applications.

 

ToneBoosters TB Sibilance v3.1.0 MAC OSX
AU - VST [Ked]

Code:

http://www41.zippyshare.com/v/Mh2LEScQ/file.html

Since my post was not published @ AudioZ I bring it here.