The principle almost nobody starts with
When a vocal is not cutting through, the instinct is to boost it. Turn it up, add some presence, maybe some air. This usually makes things slightly better and then stops working, and the reason is that the problem is rarely on the vocal track.
Two sounds occupying the same frequency range at the same moment mask each other. The louder one wins, and turning up the quieter one just makes both louder. If a guitar is filling the same 2–4 kHz region as the vocal’s consonants, no amount of boosting the vocal there will produce clarity — it will produce a louder fight.
So the first question is not “what should I do to the vocal?” It is “what else is in the vocal’s way?” Cutting 2 dB from the guitars around 3 kHz will do more for vocal clarity than any move on the vocal channel, and it will leave the vocal sounding like itself.
This is why professional mixes often have surprisingly little EQ on the lead vocal. The work was done elsewhere. It is also why our mixing engine builds a masking map first — where two sources compete, and for how much of the song, is the measurement the moves come from.
What lives where in a voice
These are orientation, not settings. Every voice is different and any number you copy from an article is a starting guess at best.
- Below 80 Hz — nothing you want. Rumble, handling noise, plosive energy, air conditioning. Almost always removed.
- 80–200 Hz — the fundamental of most singing voices, and the body of the sound. Cutting into this makes a voice thin; it is also where proximity effect piles up on a close microphone.
- 200–500 Hz — warmth, and the region that turns to mud fastest. Nearly every instrument has energy here, so it is the most competitive real estate in a mix.
- 500 Hz–1 kHz — where boxiness lives. A small cut here often opens a vocal recorded in an untreated room.
- 1–4 kHz — intelligibility. Consonants, articulation, the sense of the voice being in front. Also where the ear is most sensitive, so small moves are loud moves.
- 4–8 kHz — presence and sibilance together. This is the region you want more of and also the one that will hurt.
- Above 10 kHz — air. A gentle shelf here adds a sense of quality; a steep one adds noise and sibilance.
An order of operations
- Fix the recording first. Noise, clicks, excess room — deal with those before shaping tone, because equalising a noisy track equalises the noise too. Removing noise from a recording covers which defect wants which treatment, and in what order.
- High-pass. Remove what is not music.
- Subtractive EQ. Take out the problems: resonances, boxiness, mud.
- Compression. Now the compressor is reacting to a clean signal rather than to a 300 Hz resonance triggering it on every phrase.
- De-essing. After compression, because compression brings sibilance up.
- Additive EQ. Last, in context, and less than you think.
The order is not sacred, but this sequence removes a class of problem where each stage fights the one before it.
Subtractive moves, in practice
The high-pass filter. Somewhere between 70 and 100 Hz for most voices. Sweep it up until you hear the voice start to thin, then back off. On a deep male voice you may be as low as 60 Hz; on a bright female voice, 120 Hz can be fine. Listen rather than reaching for a preset.
Finding resonances. Take a narrow band, boost it substantially, and sweep slowly through the low mids. Frequencies where the sound becomes unpleasant or honky are the candidates. Then invert the boost into a cut, and — this matters — make the cut much smaller and much wider than the search boost was. The narrow boost is a detection tool, not a template for the fix.
Two or three decibels, with a moderate Q, at two or three places, is a normal amount of corrective EQ on a vocal. If you have eight bands doing six decibels each, something upstream is wrong — the microphone choice, the room, the performance distance. Those are the things that are hard to fix later, and they are worth fifteen minutes before the take rather than an afternoon after it.
The move people miss: carve the space on the other tracks. A gentle, wide 2 dB dip centred where the vocal lives, applied to the guitars, the keys and the synth pad, is worth more than anything you can do to the vocal itself. You can automate it to appear only when the vocal is present, so the instruments keep their full sound in the gaps.
When to boost
Boost broadly and gently. A wide shelf or a low-Q bell of 1–3 dB is almost always better than a narrow peak, because narrow boosts sound like an effect and wide ones sound like a different microphone.
- Presence — a wide lift somewhere between 2 and 5 kHz brings a vocal forward. Find the frequency by ear for the specific voice; the difference between 3 kHz and 4 kHz on the same singer is substantial.
- Air — a shelf above 10 kHz, a decibel or two. Check on headphones, since this is also where sibilance lives and headphones reveal it.
- Weight — a small lift around 150–250 Hz for a voice that sounds thin, if and only if the mix has room for it there. Usually it does not.
And always compare at matched level. A boost raises the overall level, and louder sounds better regardless of whether it is better. Match the gain before you decide, every time, or you will simply keep choosing the loudest version of everything.
Dynamic EQ and de-essing
Many vocal problems are intermittent. The voice is only harsh on the loud notes; it is only boomy on the low ones; sibilance only appears on certain words. A static EQ curve has to be set for the worst case, which means it is wrong the rest of the time.
Dynamic EQ solves this by applying the cut only when the energy in that band exceeds a threshold. A 4 dB dip at 3 kHz that engages only on the loud notes lets the quiet phrases keep their brightness.
De-essing is dynamic EQ pointed at sibilance. Two practical notes: place it after compression, since compression raises sibilance, and set the detection band as narrowly as the problem requires. A de-esser triggering across the whole top end will dull the entire vocal every time the singer says a word with an S in it.
If a single word is the problem, automate the level of that word rather than processing the whole track. Clip gain is unfashionable and it beats every plugin for surgical fixes.
Everything above is wrong in solo
A vocal that sounds beautiful soloed is frequently the wrong vocal for the mix. Soloed, you will naturally add low end for warmth and high end for detail, and both of those are in contention with the bass and the cymbals once everything is playing.
Do the surgical work in solo — finding a resonance is easier without distraction. Do all the tonal decisions in context, with the whole mix playing. Bypass frequently, at matched level. If you cannot tell whether the EQ is engaged, it is probably doing about the right amount.
And check on something small. A phone speaker has almost no low end, so it tells you immediately whether the vocal is carried by the midrange — which is where the vocal in every record you admire is actually carried. A phone is also mono, so it is the fastest check for a vocal that partly disappears when the sides collapse; the stereo width analyser shows you the same thing as a measurement.
Common mistakes
- Copying settings from a video. A move that transformed someone else’s vocal was a fix for their microphone, their room and their voice. On yours it is a random change.
- Boosting to fix masking. If the vocal is buried, look at what is burying it first.
- Comparing at different volumes. The most reliable way to make a bad decision in audio, and it never stops working.
- Too many bands. A vocal with nine EQ moves usually needed a different microphone position.
- Fixing the room with EQ. Boxiness from an untreated room is a broad resonance smeared in time, and EQ addresses only the frequency half of it. A blanket and a closer microphone do more. When the take already exists and the room is in it, dereverb treats the time half — reducing the recorded space rather than equalising around it.
Keep reading