Dynamic EQ explained

A vocal that hisses on every "s" but sits perfectly on the vowels is the textbook case for dynamic EQ, and it's the one most producers still try to solve by drawing automation curves on a static high-shelf. The shelf either dulls the take or misses the consonant. A dynamic bell at 7 kHz, threshold set just below the peaks of the sibilants, fixes it in about a minute and leaves the rest of the vocal untouched.

That gap between "draw it in by hand" and "let the band fire when the problem arrives" is the entire pitch for dynamic EQ. If you've heard the term but haven't internalised what it replaces, this is where it earns the slot.

What it does in plain signal terms

A static EQ applies a fixed gain at a given frequency, all the time. Boost 3 dB at 5 kHz and you have boosted 3 dB at 5 kHz on every sample of the file.

A dynamic EQ adds a threshold. The band sits at zero until the energy at that frequency crosses the level you set, at which point it cuts (or boosts) up to a maximum amount you also set, then releases back to zero. Below threshold, the signal is unprocessed. The filter is only doing something when the problem is present.

The comparison that trips people up is dynamic EQ against multiband compression, because both react to level inside a frequency range. The practical difference is what happens when nothing is triggering. A multiband compressor runs the signal through crossover filters whether it's compressing or not, and those crossovers colour the sound on their own. As Production Expert has noted, "when set to a null setting with no processing happening a multi-band compressor will still change the sound due to the presence of these crossover filters." A dynamic EQ leaves the signal alone until a band engages. You also get parametric Q values a multiband can't reach, and bands that overlap freely, which multibands can't do at all.

Multiband compression is still the right call for broad dynamic shaping on a bus. Dynamic EQ wins anywhere the problem is narrow and intermittent.

The three jobs it does better than the alternative

De-essing without a dedicated de-esser. A dedicated de-esser is a dynamic EQ with the controls hidden. The detection band is preset, the Q is preset, and on some plugins you can't move the centre frequency more than a hair. When the sibilance on a particular vocalist sits at 8.4 kHz and the de-esser's detector is centred at 6 kHz, you get a lispy result because the plugin is dimming a wider range than the actual problem. A dynamic bell at 8.4 kHz with Q of 4 and 4 dB of range hits only the offending band. FabFilter's own Pro-Q 4 documentation describes exactly this workflow: "you could use a narrow dynamic bell filter to suppress sibilance in a vocal recording."

Taming resonances that only appear on loud notes. Acoustic guitars often have a body resonance somewhere between 100 and 300 Hz that sits politely on fingerpicked passages and booms on hard strums. A static cut at that frequency thins out the quiet sections to fix a problem that only shows up in the loud ones. A narrow dynamic bell (Q 6 or 8) with the threshold set just above the resonance level on soft playing leaves the fingerpicking alone and pulls 3 or 4 dB out of the booming strums. Same principle on electric guitar room peaks, single-string fret buzz, and room-mode buildups on the kick.

Unmasking a vocal from a guitar without ducking the whole band. This is the sidechain trick. Put a dynamic EQ on the guitar bus, find the frequency where it competes with the vocal presence (usually somewhere between 1 and 4 kHz), and feed the vocal track into the band's external sidechain. The guitar's level at that one frequency dips when the vocal sings and returns to full when the vocal stops. Two or three dB of range is usually enough. A traditional sidechain compressor would duck the entire guitar by the same amount, which is why mixes done that way sound like the guitar is breathing around the vocal. The dynamic EQ only carves where the vocal lives.

When not to reach for it

If the problem is constant across the whole track, a static EQ is faster, cheaper on CPU, and easier to recall. A muddy 250 Hz buildup that's present in every bar doesn't need a threshold; it needs a cut.

If the problem is a one-off section change (the second verse needs less top end, full stop), that's clip gain or section automation, rather than a dynamic band. Dynamic EQ earns its slot when the problem is level-dependent and recurring.

The other trap: stacking a static cut and a dynamic cut at the same frequency. The static one handles the average and the dynamic one catches the peaks, and together they over-subtract. Pick one.

How to dial it in

Find the frequency first with a static EQ. Boost a narrow bell, sweep it across the suspect range while the problem passage plays, and stop when the problem is at its worst. That's your target. Switch the band to dynamic mode.

Set Q to match the width of the problem. Sibilance and resonances want narrow bells (Q 4 to 8). Sidechain unmasking wants something wider (Q 1.5 to 3) because vocal presence isn't a single spike.

Set the dynamic range modestly. Start at 3 or 4 dB of reduction. If the problem persists, increase before assuming the band itself is wrong. Most useful corrections land under 6 dB.

Set the threshold by watching the gain reduction meter on the band itself. Zero movement during quiet or average passages and a clear hit on every problem moment is what you're after. If the meter never returns to zero, raise the threshold; the band has become a static cut with extra steps.

Times depend on the job. De-essing wants fast attack (1 to 5 ms) to catch the consonant onset, and a release of 40 to 80 ms so the band lets go before the next syllable. Resonance taming wants attack of 5 to 20 ms and release of 80 to 200 ms. Sidechain unmasking sits in the same release range so the band doesn't chop between words.

iZotope's Sam Loose walking through dynamic EQ applications in a mix.

Where to start

If you've never used dynamic EQ on a session, put one band on your lead vocal at 7 kHz, Q of 4, 4 dB of range, attack 2 ms, release 60 ms, and set the threshold so it fires only on the loudest sibilants. That single move covers the de-esser slot, and once you've heard what it does there, the resonance and sidechain uses follow from the same controls.