Fundamentals · equalization
EQ for voice: cut the rudeness, then stop
Voice EQ, done with manners, is three moves: a high-pass filter around 80–100 Hz to evict rumble, a cut of a few decibels somewhere in the 200–400 Hz region if you sound muddy or boomy, and — only if dullness survives those cuts — a gentle lift around 2–5 kHz for presence. That is the entire starter recipe. It runs free on Windows through Equalizer APO, and the discipline that makes it work is subtraction before addition: EQ is a scalpel for what offends, not a megaphone for what pleases.
Before you touch a single band
EQ is step five of the software chain for a reason. If the wrong microphone is listening, if your gain is starved or clipping, or if an automatic "enhancement" is still meddling, EQ will merely decorate the fault. Settle those first, then record the thirty-second baseline you will A/B against — honest ears beat every preset ever shipped.
Where speech lives: a map of the dial
- High-pass at 80–100 Hzevicts rumble and desk thumps
- Cut a few dB at 200–400 Hzonly if you sound muddy or boomy
- Lift 2–3 dB around 2–5 kHzonly if dullness survives the cuts
| Region | What lives there | When it misbehaves | The correction |
|---|---|---|---|
| Below ~80–100 Hz | Rumble, desk thumps, traffic, HVAC — almost no speech | Boom, mud, meters dancing to nothing | High-pass filter. Always. This is the free lunch of voice EQ. |
| ~100–250 Hz | The body and warmth of most voices | Too much: boomy, "radio DJ in a barrel" | Small cuts only — this is also where thinness begins if you overdo it |
| ~200–400 Hz | The mud shelf | Muddy, congested, blanket-over-the-mic | Cut 2–4 dB, sweep within the range to find your voice's worst offender |
| ~500 Hz–1 kHz | Boxiness, honk | Sounding like a cardboard tube | Gentle cut if present; many voices need nothing here |
| ~2–5 kHz | Presence, consonant clarity, intelligibility | Too little: dull, distant. Too much: fatiguing | Modest lift (2–3 dB) only after the cuts; less than you think |
| ~5–8 kHz | Sibilance — the S and SH energy | Piercing esses | Small cut where the S bites; sweep to find it |
Exact frequencies vary by voice and mic; the ranges above are standard craft guidance, not measurements of yours. The sweep is the method: exaggerate a band, slide it until the ugliness peaks, then cut modestly at that spot.
The free route on Windows: Equalizer APO
Equalizer APO applies EQ system-wide at the driver level — every app receives your corrected voice without knowing anything happened. It is free and open source (GPL v2), supports Windows 7 through 11, and stood at version 1.4.2 as of 20 September 2026. We have no affiliate relationship with it; it is simply the standard answer on Windows, and it costs nothing.
- Install from the official SourceForge project (sourceforge.net/projects/equalizerapo), select your microphone as the device to process during setup (it defaults to playback devices), and reboot when asked.
- Open its Configuration Editor and build the recipe: one high-pass filter at 80–100 Hz, one peaking cut in the 200–400 Hz region (start −3 dB, bandwidth generous), and, if needed after listening, one peaking lift of 2–3 dB around 3–4 kHz.
- Record your thirty seconds, A/B against the baseline, and adjust single bands by single decibels. Manners are increments.
On macOS there is no exact system-wide equivalent baked in for input; some USB microphones ship companion software with input EQ (see who processes what), and recording apps can EQ after capture. Where no polite path exists, we say so rather than invent one.
The three sins of voice EQ
Boosting before cutting
A presence boost on top of un-cut mud gives you loud mud. Subtract the offense first; half the time the "missing brightness" was merely buried.
Correcting by numbers alone
Recipes — including ours — are starting points. Your voice, mic and room form a combination nobody else has measured. The thirty-second recording is the referee; the numbers are only the opening bid.
Expecting EQ to perform resurrection
EQ reshapes what was captured. It cannot restore treble a Bluetooth call profile never transmitted, un-mix a distant room, or invent detail a starved gain setting never caught. Fix the capture, then sculpt it.
EQ shapes the voice. Suppression removes the world.
No equalizer can subtract a keyboard or a kettle — that frequency range is your voice's range too, which is why noise removal is its own discipline. Krisp handles it across every app: 7-day full trial, no card, then $8/month billed annually on Core. Facts checked 20 September 2026.
Add Krisp after the EQ Disclosure, as manners require: this button is an affiliate link, and a subscription through it may pay the desk a fee without adding a penny to your bill. The EQ recipe above is free forever, and Equalizer APO pays us nothing.EQ questions, answered
What EQ settings make a mic sound better for talking?
Start with: high-pass at 80–100 Hz, −3 dB around 250–350 Hz, and nothing else. Listen. Add +2 dB around 3–4 kHz only if you still sound dull. That trio fixes the overwhelming majority of "my mic sounds boomy/muddy/flat" complaints, and every further band should be earned by a specific offense you can hear.
Should I EQ before or after noise suppression?
Conceptually, suppression prefers the fullest, most natural signal, and most system-wide chains put your EQ before app-level suppression by construction. In practice: keep the EQ moves modest and the order ceases to matter audibly. What matters is not running two suppressors — see the gate versus suppression etiquette.
Can EQ remove background noise?
Only noise that lives where speech does not — the high-pass genuinely removes rumble, and that is real noise reduction. Anything overlapping your voice (keyboards, chatter, fans' upper harmonics) cannot be EQ'd out without carving the voice too. That job belongs to suppression, which is a different tool with different manners.