Four EQ moves that make speech intelligible
What EQ is doing
An equaliser changes how much energy exists in each frequency band. That is all. It cannot add detail that was not recorded, it cannot remove a sound that shares frequencies with the voice, and it cannot fix a bad room.
What it can do is remove energy that is not the voice, and rebalance what remains so the consonants come through. For spoken word that is four moves, and most speech needs no more than three of them.
The human voice, by frequency
Rough map, and worth holding in your head:
- Below 80 Hz — nothing useful in a voice. Rumble, traffic, handling noise, air conditioning.
- 80–250 Hz — fundamental of most voices. Weight and body. Too much and it is boomy.
- 250–500 Hz — the "boxy" region. Room resonance lives here and cutting a little often clears a mix.
- 1–4 kHz — presence and intelligibility. Consonants live here, and this is the band that decides whether words are understood.
- 5–8 kHz — sibilance. The
sandtsounds. - Above 10 kHz — air. Pleasant, and mostly absent from cheap microphones anyway.
Move one: the high-pass filter
A high-pass filter removes everything below a set frequency. Set it at 80 Hz for a male-range voice and 100 Hz for a higher one, with a slope of 12 or 18 dB per octave.
This is the single most valuable EQ move in video work and it should be on every dialogue track by default. It removes traffic, air conditioning, footsteps through the floor, wind and handling noise, none of which is in the voice.
The test that it is set correctly: sweep it upward until the voice starts to sound thin, then back off by 20 Hz.
Move two: the notch for hum
Mains hum sits at exactly the mains frequency — 50 Hz in India, Europe, Africa; 60 Hz in the Americas — plus its harmonics at 100, 150, 200 Hz and so on.
It is narrow and it is at a known frequency, which makes it the easiest thing in audio to remove. Use a notch filter — a very narrow, deep cut — at the fundamental and the first two or three harmonics. A high-pass alone will not do it, because the harmonics are inside the voice range.
Many tools have a dedicated hum removal or de-hum module with a 50/60 Hz switch. Audacity has Notch Filter; Resolve's Fairlight has a De-Hummer.
Move three: the boxy cut
If the voice sounds like it was recorded in a cupboard, sweep a narrow boost around 300–500 Hz until the boxiness is exaggerated, find its centre, then turn that boost into a cut of 3 to 5 dB with a moderate width.
This is the standard find-it-by-boosting technique and it works for any resonance. What it cannot do is remove the reverberation itself — that is a different problem, dealt with two lessons on.
Move four: the presence lift
A gentle shelf or wide bell of 2 to 4 dB centred around 2–3 kHz raises the consonants and makes words easier to catch.
Do this last, and do it sparingly. Too much and the voice becomes harsh and fatiguing, which is worse than slightly dull. If you find yourself wanting 8 dB, the problem is that the microphone was too far away and EQ is not the fix.
De-essing, when the lift bites back
Raising presence often raises sibilance with it. A de-esser is a compressor that only responds to a narrow high band — typically 5 to 8 kHz — turning down the s sounds and leaving everything else.
Set the threshold so the gain reduction meter moves only on the sibilants, by 3 to 6 dB. If it is working constantly, the band is wrong or the threshold is too low, and the voice will sound lisping.
The order, and why
High-pass → hum notch → resonance cut → presence lift → de-ess → compression.
Subtractive moves come first because they reduce the signal reaching the compressor, and a compressor that is being triggered by 60 Hz rumble is ducking the whole voice every time somebody's chair creaks.
What EQ cannot do
Say these plainly, because people spend hours trying:
- Remove room reverb. The reflections share every frequency with the direct sound. Dedicated de-reverb tools exist and work partially; EQ does not.
- Separate two people talking at once. Both voices occupy the same bands.
- Remove wind rumble that is inside the voice range. A high-pass takes the low part; the rest is in the recording permanently.
- Fix distance. A voice recorded across a room has a poor ratio of direct to reflected sound, and no filter changes a ratio.
Free tools
Everything above exists in Resolve's Fairlight page at no cost, including the parametric EQ on every channel and the De-Esser and De-Hummer. Audacity has filter curve EQ, notch and a basic de-esser via plug-ins. Reaper's ReaEQ and ReaXComp are excellent and Reaper's evaluation is unlimited. For plug-ins, TDR Nova is a free dynamic EQ that will do the de-essing and the resonance work in one.
Today
Take a voice recording, add a high-pass, and sweep it up from 20 Hz while listening. Find the point where the voice thins, then back off. You now have the only EQ move most dialogue needs.
The one thing to keep
Four moves cover most speech — a high-pass around 80 to 100 Hz, a notch at the mains frequency and its harmonics, a small cut in the 300 to 500 Hz boxy region and a gentle lift around 2 to 3 kHz — applied subtractively before any compressor sees the signal.
Before you move on
A voice recorded in a bare room sounds hollow and distant. An editor cuts 4 dB at 400 Hz, which helps slightly, then tries progressively deeper cuts, which make the voice thin without removing the hollowness. Why?
Pick the one you would defend. Nobody sees your answer.