VocalsMixingVocal Chain Order
EQ, compression, de-esser, reverb. The order is not taste — each stage changes what the next one hears.

The range that makes a mix sound loud is the range that makes it hurt. Keep the presence and lose the fatigue.
Nothing about 2–5 kHz is inherently harsh. It is simply the region where human hearing is most sensitive, so every decibel you put there sounds like more than a decibel. Presence lives here. So does the thing that makes people turn your track off after ninety seconds.
The ear canal has a resonance that boosts this region before the signal even reaches the eardrum, which is why content around 2–5 kHz reads as louder and closer than content anywhere else at the same level.
That creates a trap. Every element sounds more present when you push it there, so vocals, snares, hats, leads and saturation all end up competing in the same narrow band. Individually each choice was right. Together they stack into pain.
Take a narrow bell, boost it hard, and sweep until it becomes genuinely unpleasant. That frequency is your problem. Note it, remove the boost, then cut a fraction of what you boosted — the sweep exaggerates on purpose.
One or two elements are almost always responsible. Cutting on the master takes presence away from everything, including the parts that were fine. Solo elements against the mix until you find the one that carries the sting.
If the harshness appears in the chorus but not the verse, a static cut is the wrong tool — it will dull the quiet sections to fix the loud ones. A dynamic cut that engages above a threshold leaves the verse intact.
Turn the mix down until it is barely audible. If it still reads as harsh, the balance is wrong and no EQ move will hide it. If it only bites when loud, you are dealing with level, not tone.
Wide cuts remove intelligibility along with the harshness — consonants, snare crack and the attack of every plucked sound all live here. A broad shelf pulled down by 4 dB will fix the fatigue and leave you with a mix that sounds like it is playing in the next room.
A resonance is narrow, so the cut should be too. One or two decibels on a wide Q, three to six on a narrow one. If you need more than that, the problem is upstream: too much saturation, or too many elements doing the same job.
Removing energy from the most sensitive part of the spectrum makes a mix immediately quieter and duller, and the reflex is to undo it. Match the levels before you judge: turn the processed version up until it is as loud as the unprocessed one, then compare.
Better still, come back the next day. Ears adapt to harshness within minutes, which is exactly why it survives the session that created it.
Press & Media
You may also like
VocalsMixingEQ, compression, de-esser, reverb. The order is not taste — each stage changes what the next one hears.
Sound designVocalsBoth make an instrument talk. One uses your mouth as the filter, the other uses your voice as a control signal.