Typical EQ discussions look at frequency cuts and boosts intended to bring out desired qualities in the voice. But sometimes there are obnoxious problems that need to be fixed. Or we want to use EQ as an effect. Sounds like a job for extreme EQ manipulation.
When a performer works too close to the mic, and/or blows a lot of air, plosives may seem more like small nuclear blasts. All of that extra bass sounds unnatural and can be distracting. A severe low rolloff (high pass) filter can often make it sound normal again, but can also make everything else sound thin and terrible. A better solution: EQ only a small portion of the recording, the part where the plosive sounds bass heavy.
In a DAW, low frequencies appear longer than high ones. Compare what you hear with what you see to help locate just the part of the recording with the explosive sounding plosive. Separate the “pop” from the rest of the voice recording. Select just that part, apply a low roll off at 150 Hz and listen to it. If that kills enough bass to sound natural again, great. But often you will need to undo, set the roll off higher and listen again. Sometimes you can stop at 200 Hz, sometimes 250 Hz, other times you need to go even higher. Steep slopes like 12 and 18dB per octave seem to work well.
Any signal processing has the opportunity to create latency. Filters often change the timing of a recording. This means that once you get the frequency and slope worked out, the processed section of voice may no longer transition smoothly to the surrounding audio. One solution is to process a larger section of audio, cut it back down to a section that fits and slide it into place. A simpler and often quicker solution is to cross fade in and out of the processed section. Experiment with these techniques in conjunction with EQing the overly plosive section of audio and your p-pop elimination kung fu will get better and better the more you try.
Here’s an sweet tutorial for removing plosives in Reaper.
When a voice is well recorded, sibilance will have already been avoided through mic selection and placement. If all “s” sounds are harsh, simple EQ may not be a good solution. But sometimes a voice sounds sibilance free most of the time, with occasional harshness. Like occasional plosives, sibilance may be made to sound more natural by selecting only the small portion that sounds bad and EQing that only.
Sweep through the 3-12k Hz region with a narrow Q and a significant boost and listen for sibilance (see Boosts for details). Or have a quick look at an RTA for an obvious peak when you hear sibilance. If you can identify a frequency to cut, doing so may return that “s” sound to something more normal (less distracting). As with plosives, aim your EQ at the part of the recording that is sibilant and leave the rest unaltered. I find a Q of 4 or narrower is needed. The cuts are pretty drastic. Start with 6dB and listen. But there are times when 9dB, 12dB, 15dB or more may be needed to to get things sounding good again.
But if every “s” sound rips out your ears, EQing each one by hand could be a huge time suck. When a recording is consistently sibilant throughout, you may need a de-esser.
A DULL ROAR
Recording on the other side of glass takes some effort. Recording underwater requires even more effort. Both of these can be approximated using extreme EQ. Set a high roll off (low pass) filter at around 8k Hz. Then adjust it lower and try different filter steepnesses until it sounds convincing. Rolling off for glass will be fairly subtle. Under water will be a lot more muffled.
THIN TO WIN
My favorite way to make a voice sound like it is coming from a phone, intercom, toy speaker, megaphone, etc. is to play the voice through the device and re-record it. Nothing is quite as convincing as the real thing. But in a pinch, many of these sounds can be approximated with extreme EQ.
Take a low roll off (high pass) filter and set it higher than you would want for a full sounding, realistic voice. You want to set it higher than 100 Hz. Somewhere just below 200 Hz things start to get weird. And they get weirder the higher you go. Decide for yourself how thin you need to make the voice sound for the effect to be convincing. For a phone you might add some distortion and/or severely lower the bit rate to really sell the idea. For an intercom, try adding a short delay or tight reverb. For a toy speaker go especially high with the roll off and more severe with distortion than for a phone. For a megaphone, make greater use of delay… either louder than you might for an intercom and/or use a longer delay time.
THE NARROW ROAD
Combine high roll off or switch to band pass filter for more vintage phone, intercom, toy, and megaphone effects. You’ll hear very narrow effects like this on some vocals during an intro, bridge, or other low fidelity moments in some songs. If you’re not trying to emulate something in the real world, experiment with really wacky settings and hear what works for your production. You can even automate frequency sweeps using band pass, high pass, or low pass filters to create a sense of motion, revelation, or closure.
The most obvious approach to EQ is push up more of what you want to hear. Getting the right microphone for the voice and recording it well should mean little or no EQ is needed. Or as my friend Khris Brown likes to say about equalizing voice: “less during initial recording.” And if you do need EQ, cut frequencies first, as previously discussed.
But you may not have the luxury of recording the voice, in which case the person who made those choices may not have set you up for success. And let’s face it, we all make mistakes and figure out later that something needs to be brought out with EQ. Don’t resist using an EQ boost based on the lofty ideal that you shouldn’t need it. Do what needs to be done to inspire listeners.
I seem to boost frequencies after compression, rather than before. Because when I EQ after, boosts work more intuitively. Boosting EQ into compression can make the compressor more responsive to the louder frequency range. And since compressors fundamentally reduce gain, boosting into compression can have the opposite effect sometimes. When I boost after compression it feels normal, natural, and obvious. Conversely I like to cut before the compressor so it doesn’t respond to undesirable stuff. Sure there are times I may break these norms, but it’s pretty rare.
Because most people record with directional microphones, proximity effect leaves many voice recordings with plenty of bass. But when a recorded voice sounds thin, I raise a parametric EQ 3 to 6 dB and sweep through the 80 to 200 Hz region listening for warmth. I tend to start with a moderately narrow Q: 2 to 4. If things sound good through a wide range then I’ll open up the Q. I’ll back off the gain until it feels warm, but not tubby. In addition to listening for warmth, I’m also listening for weird room resonances (which will overhang past the voice like a short reverb) and other, less pleasant characteristics. I don’t want to trade a solid low end for another problem. So it’s a combination of finding warmth without also accentuating something problematic.
Just north of here, in the 200-350 Hz region, things can sound thick. I don’t tend to boost here because it can make a voice sound nasal, chesty, and/or congested. I’m more likely to cut here than boost. Unless it actually sounds good (which is occasionally does), in which case I will boost it. Any audio “rule” should be considered breakable. Ultimately, how it sounds matters more than what you’re supposed to do.
The old Academy Curve had a presence boost in the 2-5k Hz range. When you need more “here and now,” this is the range to sweep. As with warmth, I like a moderately wide Q and a boost of 3 to 6 dB to go hunting for this flavor. In the lower part of this range things can honk. As you move up things can get sibilant. Find the presence frequency while avoiding honk and/or sibilance; try not to trade presence for a bigger problem.
Sometimes boosts in this range will exaggerate mouth noise. If the performer was sticky and click-y, you’ll be hard pressed to boost in the 2-5k Hz range and not bring out mouth noise in the process. In my experience EQ isn’t going to help you get rid of mouth noise, but it can definitely make it worse.
If you know the mic used to record the voice, this may influence your frequency choice for boosting. For example, the Shure SM58 tends to give a lot in the 5-6k Hz region. If the voice is miked by a 58, you probably don’t need to boost any more in that range. It’s far more likely that you need something below that to get more presence (and/or cuts are needed somewhere below 300 Hz, and/or the mic position is bad).
You may find if you work with the same person a lot you get to know characteristics of their voice. For example, I known an actor who has sibilance at about 11k Hz. So I don’t like to boost her voice in that range. Generally speaking, if a male voice is sibilant it will tend to be in the 3-5k Hz range. If a female voice is sibilant it will tend to be in the 5-8k Hz range. But that’s just a strategy. Don’t follow it like a recipe or you may inadvertently trash a recording of someone’s voice who is sibilant outside of the usual range. Sometimes you can see sibilance on a real-time analyser, which may help you find it more quickly. But if you sweep through the 3-12k Hz region with a narrow Q and a significant boost, your ears will tell you where the harshness resides. Don’t emphasize it in the name of presence. (More on de-essers in a future post.)
UP IN THE AIR
Somewhere above sibilance you start to get an airy quality. The voice and microphone play a huge role here. If the mic falls off rapidly above presence frequencies, you may boost like crazy at 12k Hz and never get much air. If the performer just doesn’t have an airy voice, EQing for it isn’t going to be as effective. Being able to boost the airy part of a voice can be elusive compared with warmth or presence.
Best practices for enhancing air are different for dialog and a vocal. In dialog folks generally don’t go for a lot of air in the voice. But in music, air in the voice can sound magical. By all means break the rules if it works for your project, just know that the expectations are different for air in the voice depending on whether you are mixing for dialog or a singing vocal.
EQ IN MATHMAGIC LAND
Tonal sounds of the voice will usually have harmonic content. So if you find something you like at 120 Hz, you will probably like what’s going on at 240 Hz, 360 Hz, 480 Hz, etc. Remember warnings earlier in this article about how some boosts can cause unwanted consequences? Well if you love the warmth at 120 Hz but there is some crazy HVAC noise it emphasizes too, try to get warmth from an interval instead.
Another use for harmonic relationships is feathering. This is where a little EQ is applied to several intervals at once instead a huge boost at only one frequency. Feathering works pretty well with voice tone, but try it with other sources too.
So when you sweep through and find something worth boosting, consider intervals of that frequency to boost instead, or in combination.
If the voice sounds great alone, but seems lacking when combined with other elements, you may need to boost something in the voice. But don’t forget that you may benefit just as much, if not more, from cuts in other elements that leave sonic landscape available to hear that great voice. You can even cut a little in competing sounds and boost a little of the same in the voice, getting the sound you want without resorting to extreme cuts or boosts.
In professional relationships I like to tell people, “I reserve the right to be wrong.” Originally this was a simple ego check; a way to let people know that I take responsibility for my errors as well as successes. Since everyone makes mistakes, please allow me the opportunity to fix mine. Let’s identify the problem — even if I am the source of it — so that we can ultimately get it right. But identifying my points of failure is taking new meaning.
I heard a piece on the radio about the Ikea Effect. In brief: when we invest our time and effort into something, we often become enamored with it. The pride of I made this threatens good decision making if we become so attached to it that we can no longer see things objectively. As a recovering shade tree mechanic, nacent “maker” and proud Do-It-Yourself-er I run the risk of being blinded by my own handywork. Does this microphone actually sound better because I modified it, or do I simply love using it because I am emotionally invested? Is my new workflow really the best choice for this project, or does my time spent creating it bias me?
Some of my most valued relationships are with people who will listen to my “elevator pitch” for something I am creating. I can get better results if I’m open to hearing the flaws instead of simply getting a pat on the back. And helping me is no small task. These people have to be willing to listen to my thoughts, bring some knowledge to the discussion, and tell me honestly about the parts that don’t jive. Inviting criticism isn’t fun. Or easy. But one of the reasons I like having conversations with other audio people is the opportunity to tease out the good from the bad. If I spent too long assembling that bookshelf, I may overlook my fairly obvious mistakes. I need folks who will ask, “Is it supposed to be like that?”
See also: Hammer Everything, an exploration of how tools influence decision making.