EQ Choices That Let a Spoken Intro Sit Cleanly Over Music

The worst feedback I ever got on a station was about a spoken intro. Not the music. The voice. A host had cut a fifty-second bit over the top of a slow soul record, and on my monitors it sounded fine. On a car radio it sounded like someone talking through a blanket with a bass drum behind them. The vocal and the bass guitar were fighting for exactly the same real estate and neither was winning.

That’s a classic EQ problem and it almost never gets solved with a big move. This is about the surgical, boring, unglamorous side of EQ: the kind that clears a lane for a voice without gutting the record underneath it.

Why the conflict is almost always in the low mids

A spoken voice, particularly a close-mic’d radio voice, carries its weight somewhere roughly between 100 and 400 Hz. The fundamental is down there, and so is a lot of the body that makes a voice sound authoritative rather than thin.

Problem is, so is a lot of the music. Kick, bass, the low end of a piano, the body of an acoustic guitar. Whoever mastered the track probably already pushed those frequencies to make the song feel full. Now you’re dropping a voice on top and asking two things to occupy one space.

You have three options and only one of them is good. You can boost the voice, which invites harshness and sibilance. You can cut the whole track, which flattens the record for everyone listening. Or you can carve a narrow notch in the music and leave the rest of it alone. That last one is the move.

The notch that does the work

Here’s the approach I settled on after too many badly balanced intros. Find where the voice actually lives, then cut that band in the music, and only the music, by a modest amount.

Narrow cuts are the tool for this. A Q of 3 or higher keeps you from affecting neighboring frequencies, which is the behavior you need if the job is pulling one problem band rather than redrawing the entire tonal picture. A wide Q here would scoop the life out of the record. A narrow one just opens a door.

How much? Start at 2 or 3 dB. That sounds small and it is, but 3 dB out of the low mids is audible as extra room for the voice. If you’re reaching for 8 dB you’ve misdiagnosed the problem, or the voice needs re-recording, or the track is just wrong for a spoken intro.

A cheap way to find the frequency

Solo the voice, sweep a boost around 150 to 400 Hz until it sounds the most obnoxious, then cut that exact band in the music. You’re hunting the frequency where the voice is loudest and most dense, because that’s the one colliding with the track.

I still do this by ear on a looped eight-second section. Tools help but the ear is the referee.

Clean first, sweeten later

The mastering approach that holds up for this: a cleaning EQ up front that takes care of what’s too loud, then, if you need it, a separate, gentler EQ later for broad musical moves. That two-stage logic comes straight from how a proper mastering chain is built, cleaning EQ first, then glue, then subtle sweetening.

For a spoken intro over music the cleaning stage is everything. You’re subtracting. Broad, small adjustments, in the spirit of mastering practice where subtlety is the whole game. The sweetening stage might add a touch of air around 10 to 12 kHz if the voice sounds dull, but that’s a finishing touch, not the fix.

The reason for splitting into two EQs isn’t superstition. A transparent, surgical EQ handles the notch cleanly without adding its own color, while an analog-modeled character EQ (a Pultec or Neve style, the kind of emulation mastrng.com lists among coloring EQs) belongs later if you want warmth. Mixing the two jobs in one plugin usually means the notch gets smeared by the character stage.

De-essing beats a high-shelf cut

Second-most-common mistake: the voice sounds slightly bright or lispy, so someone pulls a high shelf down across the whole intro. Now the voice is dull and the words lose intelligibility, which is the one thing a spoken intro cannot afford.

If the problem is “Ss” and “Ts” poking out, that’s a de-esser’s job, not an EQ’s. A de-esser turns down the offending high frequencies only when they get loud, transient by transient, instead of permanently shaving the top off the voice. The guidance to keep de-essing subtle and to double-check it only fires when you want it to is correct and I’d add: watch it on the music too, since a de-esser on a full bus will grab cymbals.

If you must EQ the sibilance instead, a very narrow high-frequency cut beats a shelf every time.

What I’d actually load

You don’t need boutique anything for this. A surgical, transparent EQ with a clean interface gets the notch done. I reach for a linear-phase EQ when I’m cutting a narrow band in mastered material, because the phase behavior stays out of the way, which is the whole reason surgical EQs like that exist (mastrng.com notes linear-phase designs are the transparent choice for fixing resonance problems). The trade-off is latency and a little pre-ringing if you push the cut hard, so keep the moves gentle and it doesn’t matter.

For the voice itself: a clean parametric to shape body, a de-esser after it. For the music: the narrow notch, nothing else. If the track needs broad tonal help, that’s a separate, earlier decision, not something to bolt on at intro time.

One thing I’d skip: any plugin marketed for “podcast voice” that applies one-size processing. The voice and the track dictate the settings, not a preset.

Check it on a bad speaker

The final test isn’t your monitors. It’s a phone speaker, a laptop, earbuds, a car. A spoken intro that sits cleanly on a good system can still collapse on a small speaker where the low mids are all you get. If the voice disappears there, your notch was too wide or too deep. Open it up a little and let the voice win by a hair.

That’s it, really. Find the collision, cut it narrow, cut it small, keep the de-esser subtle, and let the music breathe around the words instead of under them.