Best De-esser

[ramseysounds]

Different de-essers work better on different applications and different vocals. Try a few. Its really only a dynamic eq after all

[Krakatau]

Singlevocal algo is smart - smarter than simple bandpass on high frequencies so you can do anything with threshold and it detects ONLY sibilants

i assume this as being a step ahead into peculiar technology ,using AI to detect sibilants on a voice signal (that itself can be detected and partially demixed into ozone, for instance)

[ericzang]

I use Izotope RX De-esser plugin in broadband mode more than others I have. Lots of latency though. I still automate it here and there though, just as I do with any de-esser.

[Ploki]

Singlevocal algo is smart - smarter than simple bandpass on high frequencies so you can do anything with threshold and it detects ONLY sibilants

i assume this as being a step ahead into peculiar technology ,using AI to detect sibilants on a voice signal (that itself can be detected and partially demixed into ozone, for instance)

yep!

However Fab never claimed any AI or ML stuff.
just says this in the manual:

Single Vocal vs. Allround mode
When you want to de-ess a single vocal track, it's best to set the mode to Single Vocal. This enables a highly intelligent detection algorithm, which splits sibilance from non-sibilance.

Tips
In Single Vocal mode, you can lower the Treshold all the way to -INF dB. This reduces the dynamic range of the gain reduction: all sibilance will then be reduced by (roughly) the same amount (specified by Range).

And this is really how it works. It simply attenuates esses. If you distort the material a lot it can become less reliable so i reckon it's designed with mostly dry vocals in mind.
When i use it post-everything, I usually sidechain it with a dry version of the same source.

I haven't used a de-esser that was so robust and transparent and simple to use. I simply don't care to use anything else lol

Different de-essers work better on different applications and different vocals. Try a few. Its really only a dynamic eq after all

no it's really not.
- all broadband de-essers are more compressors than EQs. (EQ is really only on the sidechain)
- narrow band de-essers can have compressor topology and a multiband split (crossover), also not an EQ, and might not even have one in the sidechain. (i.e. Weiss De-esser)
- there's a few that behave like dynamic EQ, but i'd argue they're less common.

[jamcat]

The absolute best de-esser is going to be Melodyne.

[Ploki]

I don't like melodyne's sibilance detection. I generally disable it.
If you pitch something and sibilance is attached to the pitch note, you need to split it, or else it will absolutely destroy said sibilant. not great

[ramseysounds]

@ploki

A deesser cuts db at a certain frequency, generally to tame ‘harshness’, be it broadband or not

[Ploki]

@ploki

A deesser cuts db at a certain frequency, generally to tame ‘harshness’, be it broadband or not

Broadband means it cuts everything when something is present at specific frequency range (in case of a bandpass filter on the sidechain).

That doesn't tame "harshness", it attenuates the whole signal when sidechain is triggered.

Dynamic EQ cuts a specific frequency band when something is present at (usually) that band. (Either linear phase or phase response changes with levels of attenuation for minimal phase)

Multiband means there's a crossover and compressor works only on a specific band - which is again different because phase response is constant at the crossover point.

Then there are FFT "de-essers" (like soothe2) that again have some different characteristics.

And as I said, some de-essers are smarter than others (i.e. ProDS) and have a more complex detection than simply "threshold between 2-6khz" and they generally don't react to other high-frequency content that are not specifically sibilants (or sibiliant-like sounds).
And you can pick whether to cut only sibiliant range or broadband - but it doesn't affect it's detection algorithm.

[ramseysounds]

Think you’re arguing semantics but I’ll leave it. I’m out

[Ploki]

Think you’re arguing semantics but I’ll leave it. I’m out

No, you're simply wrong.
Any broadband de-esser is more of a COMPRESSOR than a Dynamic EQ.

I'll do 5:
https://www.pluginboutique.com/categories/59-De-Esser
AVA - multiband compressor
Weiss De-ess - multiband compressor
ProDS - broadband or splitband compressor (with smarter detection algorithm for sidechain vocals)
Antares - broadband compressor with sidechain hpf
SSL - broadband or splitband compressor

If you don't understand what's the difference between a dynamic EQ and a multiband compressor, that's your problem, not mine.

A dynamic EQ doesn't NOT have crossover filters and phase distortion associated with crossovers. Phase response is CONSTANT regardless of gain reduction.

On the other hand, a dynamic EQ does NOT have crossovers, but phase distortion on the Fs of the dynamic band, that changes with amount of gain reduction. Phase response is DYNAMIC.

Superficially similar, but it sounds different, looks different on a graph, and operates differently.

[ramseysounds]

Stop bitching. Of course I know the difference between a compressor and an eq. If you read my original post I said deesser, which cuts at a certain frequency, not a broadband deesser.

[Ploki]

Stop bitching. Of course I know the difference between a compressor and an eq.

Different de-essers work better on different applications and different vocals. Try a few. Its really only a dynamic eq after all

Let me guess, you meant a Broadband EQ?

[El°HYM]

Different de-essers work better on different applications and different vocals. Try a few. Its really only a dynamic eq after all

I mainly use De-essers on Drones & Pads to reduce some of the Harshness in the upper Frequencies. The use of spectral processing mostly works pretty fine for this Task.

[Bodhisan]

...or Waves Omnischeps where you can even have 2 de-essers at the same time...

Having vocals in all of my songs, for me, this has been the best out of all de-essers I've used - and believe me, I've tried many of these mentioned in this thread. If your sibilance is really bad, though, most de-essers, to be effective, really hurt your high end on your vocals - not a good compromise.

That said, if this is for yourself, what I've found best is be aware how you sing your s's, or turning away for the mic as you sing a word with a sibilance - the more distance from a mic, the better. I re-track if my sibilance is not acceptable. If not, the second best de-esser is using the "selection based processing" (in Logic - don't know what it's called in other DAWS) on each offending sibilance and using either your de-esser or an e.q. - or, in my case, a gain reduction of about 6 dbs works 9 times out of 10. It takes about two minutes to tame a vocal track's sibilance this way, which only affects the sibilance part of the track, nothing else gets highs tamed, as opposed to using a standard de-esser.

I used to be satisfied using Nova as a dynamic e.q. around the 4.2k area, but, again, to be effective with my worst sibilance, it had to kick in constantly with even the most minor of offenders, dampening my high e.q. When I a/b'd it, I wasn't happy with the compromise. But all this is just me and my vocals - some vocals might sound perfect for the track with what ends up being a lpf. To each their own.

[trackbout]

I don't like melodyne's sibilance detection. I generally disable it.
If you pitch something and sibilance is attached to the pitch note, you need to split it, or else it will absolutely destroy said sibilant. not great

That was how it worked before Melodyne 5. After v5, Melodyne automatically isolates the sibilant from the rest of the "blob", so when you change the pitch or the note length, the sibilant is completely unaffected. The Melodyne 5 sibilance tool absolutely smokes any traditional de-esser I've tried.