3. Hey, How are Those High-Mids Doing?
A lot of times what we perceive as trebly harshness occupies what we'd technically call the high-midrange—anything between, say, 2 kHz and 8 kHz. Sibilance can sometimes occupy this space, as can nasty resonances in crash cymbals, clangy aspects of a guitar, or the needly nonsense of a synth. Taming the high-mids can be problematic for EQ, as a static cut at these frequencies could result in a lifeless timbre overall.
A dynamic process, on the other hand, might accomplish two functions: as mentioned above, it'll only clamp down when a signal below your designated threshold tells it to. Also, you'll be able to sculpt the attack and release, so you can let the initial portion of that harsh transient in. Then you can soften its lasting impact as the compressor suppresses the nasties. This application of multiband can often mitigate the harshness while preserving the sound's innate character.
Three different types of processors are suited to handle the task: a multiband compressor, a dynamic EQ, or a de-esser in split-band mode. If you have a dynamic equalizer on hand, and you’re new to this technique, I recommend selecting that option first, as you won't have to worry about the other bands; sure, it's easy enough to put them in bypass, but why go through the extra step if you don't have to? Plus, this way you won't be tempted to use them.
Say you've got a vocal particularly resonant at 7 kHz. Here's how it would go: you'd grab a dynamic EQ, set up a 1 dB to 3 dB cut at 7 kHz with a moderate Q. Next, you’d keep the attack just slow enough to let transients through, all while using a fast to medium-fast release to recover quickly. Threshold you must set to taste. I keep the ratio—if it's offered—set relatively low. The trick is to keep the ratio or range value high enough to mitigate the problem, but low enough not to call attention to itself.
If you're using a de-esser, you can listen to the sidechain monitor to isolate the frequencies you hate and then slap it on as you would on a vocal. Sometimes a de-esser’s pre-programmed algorithm for handling this problem can actually achieve better (or faster) results than a traditional multiband compressor or a dynamic EQ. Sometimes it can’t; you’ll have to experiment until your guesses are no longer guesses.
4. Try the Ol’ Mult and Split
This is a dicey proposition, as it could have adverse effects on coherence; indeed, it's possible this trick could smear the source, so I wouldn't go about employing this technique on an element that needs to retain its innate specificity; this probably isn’t your go-to trick for the snare drum, in other words. Still, it’s effective on stereophonic vocals, guitars, synths, or any sound source you're planning on modulating with spatial and modulation effects.
In essence, you duplicate the element to a new track, low-pass the horridness out on the original channel with a gentle slope, high-pass the duplicate to about 8 kHz or so, and then apply modulation, spatial effects, or ambiance processing until the two tracks gel into a solid product.
Envision this scenario: you're working on a pop or electronic track where the vocal will receive stereophonic treatment anyway—the kind of thing an Eventide would handle in the analog world, or Ozone Imager would handle in the digital realm. Unfortunately, this vocal has a nasty resonance in the highs, and the previous techniques aren't working. Here you could mult the track, high pass the vocal to around 5 kHz or 6 kHz (carefully—with a gentle slope), low pass the duplicate to 9 kHz, de-ess it for harshness, put a little stereo chorus on the mult to widen it out a bit, and blend it back in.
Will this sound natural? No. Will it smear? Probably. But will it work within certain electronic or ethereal contexts? If done right, then yes.
Here's another hypothetical: say you've got a stack of vocals that comprise the lead hook, but the main vocal track sounds nasty in the trebles. Now we're in greater luck, as we can do our mult trick, blend the lead vocalist in with the stack, and slap an imager on the high-passed duplicate to widen those sparkles, spreading them around the stack. In this way we could, for a specific arrangement, take a liability and turn it into an asset.
5. Yin that Yang!
It has been observed that there's a symbiotic, yin-and-yang effect between high and low frequencies. Cut something in the low end, and you'll feel more presence in the highs (the opposite is also true). You can take advantage of this, using frequency bumps in the low midrange to offset or mask the harshness peaking through the highs. Be careful in doing so—you don't want to introduce tubbiness.
In my experience, this technique works better in naturalistic genres like classical and jazz, where you need to retain the character of interplay between instruments miked in a live setting. Here, only the most subtle tricks will work. In more electronic and electric genres, where frequencies desire more sculpting, you might introduce the aforementioned tubbiness, and this might not sound good. As always, let your ears guide you.
6. Try a Little Harmonic Distortion
This trick has quickly become my go-to since I figured it out (I'm sure others have figured it out first, but either way, you're sure to get some mileage out of it.) Here, we’ll use harmonic distortion, excitement, or a hardware “color box” to warm up the sound. I approach the technique as follows:
Dip the offending frequencies with a colorless, surgical EQ. You should have a lifeless track at this point—one devoid of the harshness, but still, pretty dull. Now, apply harmonic excitement/distortion of some kind. Here you can experiment with exciters brought to you by iZotope (such as Ozone, Neutron, or Trash for downright distortion), or an outboard piece of equipment. If you're using a multiband exciter, give some love to the low midrange, preferably in an area whose third-order harmonics would enhance/complement the area you've cut (use a frequency chart if you'd like to quickly calculate these harmonics). Don't go overboard here on the saturation; the aim of the game is subtle coloration, not screaming distortion.
Now, after this processing, put on another surgical EQ and restore the frequencies you cut originally. Here I like to simply duplicate the original EQ and change all my negative values to positive, so that a cut of -3.5 dB becomes +3.5 dB, for example.
If all goes well, the track should sound much like it did before, minus the aggressive harshness in the highs, giving you something more malleable to work with as a starting point.
7. Are You Using Too Many Plug-ins?
I realize the irony of putting this tip right after one that directed you to use three plug-ins, but here we go anyway:
Yes, it's happened to all of us—especially when we're starting out: we pile on plug-in after plug-in, and after awhile, we begin to wonder, "Where's that horrible harshness coming from?" Sometimes, we pile too much processing on a track, and the effect of all these color compressors rubbing up against harmonic exciters and vintage EQs is not pleasant.
There’s more than a few reasons for cold relations between different algorithms: perhaps you’re using a really old plug-in that doesn't utilize floating point processing, and you’re overloading it unknowingly. Maybe you’re undoing one processor’s effect with another that isn’t so complementary. Often coloration plug-ins sport inherent sonic characteristics, which means that anything you drive into them is further amplified within the matrix of its favored frequencies.
Here's a tip: if you're wondering why your track sounds harsh, and you've got more than three plug-ins on the track, try bypassing all of them and seeing if the harshness goes away. If it does, that's your clue right there. Next, start at the top: put on plug-ins one by one and notice at which point the harshness arrives. For this tip you have my permission in solo—just don't spend a long time listening to the track all by its lonesome, or if you do, take a break for a moment.
8. A Plug for a Plug
Yes, this is a plug, but it’s not shameless, for these plug-ins do work. I’d mention them in articles regardless of whether or not they appeared under iZotope’s masthead—and I do so regularly. Two processors in particular excel at taming high-frequency harshness, though in rather different ways.
The first is RX, whose combination of easy-to-read spectral editing and host of useful modules come in handy for spotting troublesome frequencies, and consequently, allow you to attenuate troublemakers with minimal risk to other bands. The Gain module and Magic Wand tool, in adroit hands, can target harshness at specific points and tamp it down to your exact requirements.
Of course, there's also the Spectral Repair module, which, in Attenuate mode, can make demonstrable differences. You can train this module to observe better frequency-handling surrounding a selected region. Then, it attenuates the selection based on the behavior it observes.