De-esser
A compressor aimed only at the harsh 's' and 't' frequencies, so sibilant vocals stop hissing without dulling the whole voice.
See it
What it is
A de-esser is a compressor that only listens to one narrow band, typically somewhere between 4 kHz and 9 kHz, and ducks the whole signal (or just that band) when sibilance spikes there. The result: the 's', 'sh', 'z', and 't' sounds stop stabbing while the rest of the voice is untouched.
Where it goes in the chain depends on the source, not on a rule. If the raw take is already spitty, put it early so the compressor is not triggering off esses. If a presence boost around 6 kHz sharpened the voice and the esses with it, or a fast compressor pulled the body down and left the consonants proud, put it after that stage instead. Plenty of engineers do a little in both places. Neither EQ nor compression creates sibilance on its own; they just change how much of it you notice. Find the offending frequency by soloing the detector band and sweeping until the hiss is loudest, then pull 3 to 6 dB on the worst syllables.
Gotcha: over-de-essing gives you a lisp. If 'sister' turns into 'thithter', back off the threshold or narrow the band. Also, sibilance is often a mic problem first: move off-axis by 15 degrees or back off a few inches and half the work disappears.
Ask AI for it
De-ess this vocal track: detect sibilance in a narrow band between 5 kHz and 8 kHz and apply split-band gain reduction of 3 to 6 dB only on the sibilant syllables, with fast attack and roughly 40 ms release. Leave everything below 4 kHz completely untouched, and stop short of any lisping or dulling of the consonants. Try it both before and after the compressor and keep whichever position sounds more natural on this particular voice.