How to Fix Harsh Vocals (2-5 kHz Explained)
Vocal mixing · 4 min read
Harshness in a vocal is almost never a whole region. It is one resonance, usually between 2 and 5 kHz, and the fix is a narrow cut of 3–5 dB at Q 5–8 on exactly the right frequency. A wide 1 dB cut across the whole area removes presence and leaves the harshness where it started.
Harshness lives at
2–5 kHz
One resonance · cut 3–5 dB at Q 5–8
The 60-second answer
Boost a bell +10 dB at Q 8 and sweep 1.8–5 kHz until one spot makes you flinch. Flip that boost into a cut of 3–5 dB at the same frequency and Q. Make it dynamic so it only pulls on the loud notes, then check the octave above for a second, smaller peak.
Where the resonance comes from
Three usual suspects. Most budget large-diaphragm condensers have a deliberate presence lift of 4–6 dB somewhere between 4 and 8 kHz, and singing 10 cm from one stacks that onto a voice that already peaks nearby. Rooms add their own: a hard surface within a metre of the mic sends a delayed copy back, and the comb filtering parks peaks a few kHz apart. And layers multiply it — four takes with the same 3 kHz peak sum into something nobody would accept on one take.
Why narrow beats wide
A resonance is a spike maybe a third of an octave wide. A wide, gentle cut across 2–5 kHz takes 1 dB off the spike and 1 dB off everything else, so you lose consonant clarity, the vocal slides behind the beat, and it is still sharp. A narrow cut takes 4 dB off the spike and almost nothing off its neighbours. Wide and gentle is for tone shaping. Narrow and deep is for problems.
Fixing harshness, step by step
| # | Do this |
|---|---|
| 01 |
Check that it is the vocal Mute the vocal and listen to the beat alone. Hats, snare tops and bright synths share the window, so harshness in a mix is often the sum. Mute test first |
| 02 |
Sweep with a narrow boost Bell, +10 dB, Q 8, moving slowly through 1.8–5 kHz on a busy phrase. Find the one spot that makes you pull away from the speakers. Bell +10 dB, Q 8 |
| 03 |
Turn the boost into a cut Same frequency, same Q, gain from +10 to -4. Start at -3 dB and only go past -5 dB if it still hurts. -3 to -5 dB, Q 5–8 |
| 04 |
Make it dynamic Harshness mostly shows up on loud and belted notes. A dynamic band with a range of -4 dB leaves the quiet lines alone. Range -4 dB, release 60–100 ms |
| 05 |
Look one octave up A 2.4 kHz resonance often has a friend near 4.8 kHz. A second cut of 2 dB there does more than deepening the first. Second cut: -2 dB |
| 06 |
Level-match and turn the monitors down Your ears peak in sensitivity around 3 kHz, and it gets worse the louder you listen. Judge at conversation level. Monitor at 70–75 dB SPL |
The trick: one narrow cut, checked with a level-matched bypass
Here is the test that tells you whether you found the resonance or just made the vocal darker. Set your cut, bypass the band, and trim the levels so the bypassed version is not louder. Now switch back and forth while a busy phrase loops. If you found the resonance, the cut version sounds the same brightness and less painful — you still hear every consonant and the vocal sits in the same place in the mix. If it sounds duller or further back, the frequency is wrong or the Q is too wide, so undo it and sweep again. Producers skip this test because 3 dB of extra level beats 3 dB of tone in any blind comparison, and that is how vocals end up with four overlapping cuts that each seemed like an improvement.
Stop guessing the chain.
The Vocal Chain Bible is 88 real vocal chains from Grammy-winning engineers — every plugin, every setting, in order, with free alternatives for each one. 5+ years, 100+ interviews, one PDF.
Get the Vocal Chain Bible — $48 →Mistakes that kill it
- A wide shelf cut at 3 kHz. It fixes the pain and takes the intelligibility with it. Two weeks later you are boosting 3 kHz to get the words back.
- Adding air to balance the cut. A 12 kHz boost on top of a harsh 3 kHz peak gives you a bright harsh vocal instead of a harsh one.
- Sweeping at high volume. Everything hurts at 90 dB SPL, so you will cut roughly twice as much as the mix actually needs.
- Forgetting the exciter you added earlier. Saturation and exciters build energy in exactly this region. Bypass them before you decide the voice is harsh.
Keep going
FAQ
What frequency makes vocals harsh?
Usually one narrow resonance between 2 and 5 kHz. Above 5 kHz you are into sibilance, which needs a de-esser instead.
How much should I cut for harshness?
3–5 dB at Q 5–8. If 5 dB is not enough, look for a second resonance an octave up rather than deepening the first cut.
Why do my vocals sound harsh only in the mix?
The hats, snare top and synths are stacking with the vocal in the same 2–5 kHz window. Cut the resonance, then dip the instrumental bus 1–2 dB there.
Who's behind this
Ivan Meshcheriakov · IvanFromRnD
5+ years ago I moved to LA with zero connections and started building from nothing — 14-hour days, meeting people and interning at studios. It took me 5+ years to meet and interview 100+ Grammy-winning engineers — then I turned it all into the Vocal Chain Bible.
Read my full story →