EQ for Voice Over and Podcasts: Settings That Actually Work

The short answer: for a spoken voice, start with a high-pass filter around 80 Hz (higher for higher voices), make a narrow cut wherever mud or boxiness builds between roughly 200 and 800 Hz, add a small, wide boost around 2 to 5 kHz only if the voice sounds distant, and control harsh S sounds with a de-esser rather than a fixed EQ cut. Cut before you boost, keep most moves under 3 or 4 dB, and if a problem needs more than about 6 dB of EQ, fix the recording instead.

That paragraph is the whole method. The rest of this Guide is the reasoning behind each number, so you can adapt it to your own mic, room and voice rather than copying a preset that was made for someone else's. It is written for voice actors delivering finished files to clients and for podcasters cleaning up hosts and guests, because the physics of a speaking voice is the same in both jobs. Where the two differ, mostly in how much processing a listener will tolerate, it says so.

The Frequency Map of the Human Voice

Everything an equalizer does is described in hertz, so the first job is knowing where a voice actually lives. The pitch you hear when someone speaks is the fundamental frequency of their vocal folds. In Ingo Titze's standard text on voice science, adult male speaking voices vibrate at roughly 85 to 180 Hz and adult female voices at roughly 165 to 255 Hz. Those are averages across many speakers, and plenty of real voices sit outside them, which is why this Guide talks about deeper and higher voices rather than men and women. Above the fundamental sit the harmonics and resonances that give a voice its colour, and above those sit the consonants that make it understandable.

The most useful single fact on that map comes from the microphone maker DPA, whose guide to speech intelligibility describes the band around 2 kHz as the most important for how clearly speech is understood, and the wider 1 to 4 kHz range as of high importance. Consonants live mostly in the 2 to 4 kHz region, while vowel energy usually falls away quickly above about 1 kHz. DPA also reports that filtering away everything below 500 Hz reduces intelligibility by only about 5 percent, while cutting everything above 1 kHz drops it below 40 percent. Read that the right way round: the low end of a voice carries most of its weight and warmth but very little of its meaning, and the upper midrange carries the meaning. That is why a voice that sounds big and rich in your headphones can still be hard to follow on a phone speaker, and why most voice EQ is about protecting 1 to 4 kHz from things that mask it.

Two more zones complete the map. Sibilance, the hiss of S, SH and soft C sounds, usually concentrates somewhere between 4 and 10 kHz according to Sound On Sound's guide to de-essing, and its exact centre varies a lot from voice to voice. Above 10 kHz is what engineers call air: breath, texture and a sense of openness, useful in small doses on a quiet recording and a hiss amplifier on a noisy one.

Subtractive vs Additive EQ: Cut First

There are two ways to make a frequency range stand out: turn it up, or turn down whatever is around it. Subtractive EQ, cutting, is the safer default for voice for three practical reasons. A cut removes something that was already a problem, such as room boom or a resonance, while a boost raises everything in that range, including room noise and hiss. Cuts do not eat into your headroom, so you are less likely to push a file toward clipping before it reaches the compressor or limiter. And ears adapt to boosts: after a few minutes a lifted presence band sounds normal, so you add a little more, and the file gets harsher without you noticing.

Additive EQ still has a place. A small, wide boost in the presence range can bring a slightly distant voice forward, and a gentle high shelf can open up a dark dynamic mic. The working rule is simple: cut narrow to remove problems, boost wide to shape tone, and do the cuts first so you are not boosting a problem along with the voice.

Q and Bandwidth, Explained Properly

Every bell-shaped EQ band has three controls: frequency, gain and width. Width is shown either as Q or as bandwidth in octaves, and the two run in opposite directions: a higher Q means a narrower band. Bandwidth is measured between the points either side of the centre where the filter is 3 dB from its peak, and the conversion between Q and octaves is not a simple reciprocal. RaneNote 170, the reference note on the subject, gives the formulas; the graphic below applies them to a 6 dB cut centred on 300 Hz, a typical spot for mud.

Those three numbers are worth memorising. Q 0.7 is a broad, gentle tone control that affects nearly two octaves; at 300 Hz it reaches from below the fundamental of most voices up into the low midrange. Q 1.4 is one octave and a good default for a mud cut. Q 4.3 is a third of an octave, the width of the bands on a classic 31-band graphic EQ, and the right tool for a single resonance from a small room. Adobe's documentation for Audition notes that Q values from 1 to 10 suit most equalization work, which matches how engineers use them: almost nothing on a voice needs a Q above 10, and a Q that narrow on a boost will ring.

The High-Pass Filter: Where to Set It

A high-pass filter, also called a low-cut, removes everything below a set frequency and leaves the rest alone. Below about 80 Hz there is almost nothing useful in a speaking voice: what lives there is HVAC rumble, traffic, desk and stand vibration, and the low thump of plosives that got past the pop filter. Removing it makes the file cleaner, stops that energy from triggering your compressor, and costs the voice nothing you would miss. If you only make one EQ move on a voice, make this one.

Choosing the frequency

Set the filter below the lowest fundamental the speaker actually uses. For a deep voice that usually means 60 to 80 Hz; for a higher voice, 90 to 110 Hz is often transparent. The practical method is to start at 60 Hz, raise the frequency slowly while the voice plays, and stop at the point where the voice begins to sound thinner, then back off by 10 or 20 Hz. Remember that Titze's ranges are averages and pitch drops at the ends of phrases, so a voice that averages 120 Hz may dip toward 90 Hz on a downward inflection. Filtering too high is one of the most common ways to make a voice sound small.

Choosing the slope, and the proximity effect

The slope, in decibels per octave, sets how quickly the filter cuts below its corner. Twelve dB per octave is a gentle, natural-sounding default for voice; 18 or 24 is steeper and useful when there is a lot of rumble right below the voice. One detail from Sound On Sound's Mike Senior is worth knowing if you use Reaper: narrowing the bandwidth on ReaEQ's high-pass does not make the slope steeper, it adds a resonant bump at the corner. Stack two filters if you need a steeper cut.

Close-miking adds its own low end. Directional microphones boost bass as the source gets closer, the proximity effect, and Shure's own user guides describe it as a progressive bass boost as the mic moves nearer the mouth. Some broadcast mics build a filter in for this reason: the Shure SM7B has a bass rolloff switch, which Shure suggests for reducing hum from air conditioning or traffic, and a presence boost switch for a brighter midrange. If your mic has a switch like that, decide on it at the recording stage, then set the EQ high-pass with that in place.

Finding Mud and Boxiness

Mud is a thick, woolly buildup that makes words blur together, usually somewhere between 150 and 350 Hz. Boxiness is the hollow, cardboard sound of a small room, usually higher, somewhere between 400 Hz and 1 kHz. Both are often caused by the room rather than the voice: small rooms with parallel walls reinforce particular low-mid frequencies, which is why treating the room fixes the cause while EQ only manages the symptom. Because the voice's own warmth lives in the same range, the cut has to be narrow and targeted. The standard technique is the sweep:

1. Make a band deliberately bad. Set one bell band to a boost of 8 to 10 dB with a fairly narrow Q, around 4. Turn your monitoring down first.
2. Sweep it slowly from about 150 Hz up to 1 kHz while the voice plays. Most frequencies will just sound louder. One or two will sound obviously worse: boomier, hollower, more like a cardboard box.
3. Stop where it sounds worst. That is the resonance. Note the frequency.
4. Flip the boost into a cut of 2 to 4 dB at that frequency. Widen the Q a little, toward 2, if the problem covers a broader area.
5. Bypass and compare at matched level. The voice should sound clearer, not thinner. If it sounds thinner, the cut is too deep or too wide.

Two cuts are normal; four or five usually means you are chasing something EQ cannot fix. If the same frequency turns up in every file you record, it belongs to your room, and you can save that cut as a starting preset for that room and that mic. That is the one kind of preset worth keeping, because you made it from your own space.

Presence and Clarity: 2 to 5 kHz

If a voice sounds distant or dull after the cleanup cuts, a small, wide lift in the upper midrange is the usual cure. This is the zone DPA identifies as carrying intelligibility, and it is also where a raised voice naturally gains energy: DPA notes that speaking louder shifts the spectral emphasis one to two octaves higher, which is part of why projected reads sound more present. Try 1 to 2 dB with a Q around 0.7 to 1, centred around 3 kHz for a deeper voice and a little higher for a higher voice. Then stop. This range fatigues the ear quickly, and over a 40-minute episode or a 10-hour audiobook a harsh upper midrange is far more tiring than a voice that is very slightly soft.

Air, a high shelf around 10 to 12 kHz, is the last and most optional move. On a clean recording made with a dynamic mic, a decibel or two can add openness. On a recording with any hiss, it raises the hiss by the same amount, so only use it on a quiet file and check the gaps between sentences after you apply it.

Sibilance and De-Essing

Sibilance is the one voice problem that EQ alone handles badly. A fixed cut in the 5 to 9 kHz range lowers every S along with the brightness and detail of every other word, so the voice gets dull and the worst esses still poke out. A de-esser is a compressor that listens only to the sibilant band and turns it down only when an S actually spikes. Sound On Sound's de-essing guide puts sibilant energy at roughly 4 to 10 kHz, and the SPL De-Esser, a well-known design reviewed by the magazine, uses a male/female switch that moves its centre between 6 and 7 kHz. That is a useful starting point: begin near 6 kHz for a deeper voice and 7 kHz or a little higher for a higher voice, then fine-tune by ear.

Adobe's documentation for Audition's DeEsser describes the controls most de-essers share: a threshold above which reduction starts, a centre frequency at the point where sibilance is most intense, a bandwidth for how much of the spectrum triggers it, and a choice between Broadband mode, which turns down the whole signal, and Multiband mode, which turns down only the sibilant band. Adobe recommends Multiband for most material. Its Output Sibilance Only option lets you hear exactly what the effect is catching, which is the fastest way to set the centre frequency. Aim for 3 to 6 dB of reduction on the loudest esses and none at all on normal speech. Too much and the speaker starts to lisp, a sound listeners notice faster than a slightly bright S. For mouth noise and plosives alongside sibilance, see fixing mouth clicks, plosives and sibilance.

Starting Points for Deeper and Higher Voices

Because a higher voice has its fundamental and its resonances higher up the spectrum, most of its EQ points shift upward too. The table below gives starting settings by symptom, with one column for deeper voices and one for higher voices. These are where to begin listening, not presets to apply blind, and every number moves with the mic, the room and how close the speaker sits.

For podcasts with several people, EQ each voice on its own track. A co-host with a deep voice and a guest with a higher one need different high-pass points and different mud cuts, and a single EQ on the mix bus will always be wrong for one of them. Matching two voices is mostly about matching their low end and their presence; once the high-pass and the mud cuts are right for each person, the voices usually sit together without much more work.

EQ Before or After Compression?

There is no single correct order, and Sound On Sound's Mike Senior has said as much: for subtle EQ settings, he argues, it matters little which comes first. What does matter is low-frequency energy. Rumble, plosive thumps and mud are big in level, so if they reach the compressor first they drive it, and the compressor turns the whole voice down every time a truck goes by. That is why corrective EQ goes before compression and tonal EQ often goes after. A reliable chain for spoken voice is:

1. Noise reduction, if needed. Clean the noise first so later processing does not raise it.
2. Corrective EQ: the high-pass and the narrow mud or boxiness cuts.
3. De-esser. Before the compressor, so the compressor does not react to esses; some engineers add a second, light de-esser at the end.
4. Compressor. Evens out the level of the read. See noise gates and compressors for voice for settings.
5. Tonal EQ: any presence lift or air shelf, now that the dynamics are settled.
6. Loudness and limiting. Normalise to the delivery target and set the true-peak ceiling last.

If you boost presence after the compressor, listen to the sibilance again afterwards: a presence lift raises the lower edge of the sibilant band, and a de-esser that was perfectly set before the boost may now need a slightly lower threshold.

Step by Step in Audacity, Reaper and Audition

The method is the same in every editor. Only the menus change. Apply EQ to a copy of the file or use a non-destructive effect, so you can come back and adjust it.

Audacity

1. High-pass: select the whole track, then Effect > EQ and Filters > High-Pass Filter. Set the frequency to 80 Hz (or the value you found by ear) and the roll-off to 12 dB per octave. The manual defines the frequency as the point where the signal is already 3 dB down.
2. Mud and boxiness: Effect > EQ and Filters > Filter Curve EQ. Click on the curve to add points and drag a narrow dip at the frequency you found by sweeping. Use the Graphic EQ (same menu) if you prefer fixed sliders.
3. The speech preset: Filter Curve EQ has a factory preset called Low rolloff for speech, a rumble filter long recommended on the Audacity forum for voice work. It is a reasonable alternative to the High-Pass Filter; use one or the other, not both.
4. Presence: in Filter Curve EQ, draw a broad, low hill of 1 to 2 dB around 3 kHz. Keep it wide; a sharp peak sounds harsh.
5. Sibilance: Audacity's built-in EQ cannot react only to esses, so treat the worst ones individually or use a dedicated de-esser plug-in loaded through Audacity's plug-in support.

Reaper (ReaEQ)

1. Insert ReaEQ on the voice track's FX chain. It opens with four bands, a low shelf, two bell bands and a high shelf, and shows bandwidth in octaves.
2. High-pass: change band 1's type to High Pass and set it to 80 Hz. For a steeper cut, add a second High Pass band at the same frequency rather than narrowing the bandwidth, which only adds resonance at the corner.
3. Sweep for mud: set band 2 to Band, gain +9 dB, bandwidth about 0.3 octaves (a Q of roughly 4.3). Drag it across 150 Hz to 1 kHz during playback, then turn the gain to minus 2 to 4 dB at the worst spot and widen to about 0.7 octaves.
4. Presence: set band 3 to Band, about 3 kHz, +1 to 2 dB, bandwidth 1.5 to 2 octaves.
5. Sibilance: use ReaXcomp, Reaper's multiband compressor, with one band covering roughly 5 to 9 kHz and the other bands set so they do not compress, or any de-esser plug-in. Place it before the compressor in the chain.

Adobe Audition

1. Open Effects > Filter and EQ > Parametric Equalizer. Leave the band width setting on Constant Q, which Adobe describes as the most common setting.
2. High-pass: switch on the HP band, set it to 80 Hz and choose a slope of 12 or 18 dB per octave.
3. Mud and boxiness: use a numbered band. Raise its gain, narrow the Q to about 4, sweep, then turn it into a 2 to 4 dB cut at the worst point and relax the Q toward 2.
4. Presence: another band around 3 kHz, +1 to 2 dB, Q about 0.7 to 1.
5. Sibilance: Effects > Amplitude and Compression > DeEsser. Choose Multiband, tick Output Sibilance Only while you set the centre frequency, then untick it and lower the threshold until the loudest esses lose 3 to 6 dB on the Gain Reduction meter.
6. Save your result as a preset named for your mic and room, so the next session starts from your own space rather than from zero.

The Most Common EQ Mistakes on Voice

• Filtering too high. A high-pass at 150 Hz on a deep voice removes the chest of the voice along with the rumble. Find the point where it thins, then back off.
• Wide cuts in the low mids. A broad cut from 200 to 500 Hz kills the warmth of a voice. Cut only the resonance you found, narrowly.
• Boosting presence to fix a distant mic. If the speaker was too far from the mic, the problem is room sound, and boosting 3 kHz boosts the room too. Move the mic closer next time.
• Using EQ for sibilance. A static cut dulls everything. Use a de-esser.
• Judging while louder. An EQ setting that raises level always wins a quick comparison. Match levels before deciding.
• EQing on one pair of headphones only. Check on earbuds or a phone speaker too, because that is where many listeners will hear you.
• Trying to EQ a bad recording into a good one. A clipped or very quiet take cannot be repaired with EQ; boosting a quiet file raises its noise floor right where you boost. Get the gain staging right at the source.

A Worked Example: Two Podcast Hosts

Take a common setup: two hosts in the same spare room, each on a dynamic broadcast mic a hand's width from the mouth. Host A has a deep voice; host B speaks noticeably higher. Here is how the starting settings translate into real decisions, one track at a time.

Host A gets a high-pass that starts at 60 Hz. Raising it during playback, the voice still sounds full at 80 Hz and starts to thin around 100 Hz, so the filter settles at 80 Hz, 12 dB per octave. Sweeping a narrow boost finds a boomy spot near 200 Hz, partly from the proximity effect at that close distance, and a 3 dB cut at Q 2 clears it. Host B's high-pass can go to 100 Hz without any audible loss. Host B's sweep finds no particular mud but a hollow spot near 600 Hz; since the same frequency also shows up faintly on host A, it is the room, so both tracks get a 2 dB cut there at Q 4.

Neither host needs a presence boost, because a dynamic mic close to the mouth is already present. Both get a de-esser, centred near 6 kHz for host A and near 7.5 kHz for host B, then the compressor, then loudness normalisation to the podcast loudness target. Total EQ: two filters, three cuts, no boosts. That is typical, and it is why good voice EQ is usually invisible: nobody listening notices it, they just notice that the voices are easy to follow.

Frequently asked questions

What are the best EQ settings for voice over?

There is no universal best, but a reliable start is a high-pass at 70 to 110 Hz depending on how deep the voice is, a 2 to 4 dB narrow cut wherever a sweep finds mud or boxiness between 150 Hz and 1 kHz, an optional 1 to 2 dB wide lift around 3 kHz, and a de-esser near 6 to 7 kHz. Set each frequency by ear on your own recording.

What frequency should a high-pass filter be for voice?

Usually 70 to 80 Hz for a deep voice and 90 to 110 Hz for a higher one. Start low, raise it until the voice begins to thin, then back off slightly. A slope of 12 dB per octave sounds natural; use 18 or 24 if there is heavy rumble just below the voice.

Should EQ go before or after compression on a voice?

Corrective EQ, the high-pass and the mud cuts, goes before the compressor so low-frequency energy does not drive it. Tonal boosts can go after. For subtle settings the order matters little; for rumble and plosives it matters a lot.

Is EQ different for male and female voices?

The method is the same; the frequencies shift. Average female speaking pitch sits well above average male pitch, so the high-pass, mud cut and de-esser centre usually sit higher for a higher voice. Set them by the voice in front of you, not by the speaker's gender, because many voices sit outside the averages.

Do podcasts need different EQ from voice over?

The frequencies are the same. The difference is in context: podcasts often have several voices that need to match each other, and listeners hear them for an hour on earbuds, so restraint in the 2 to 5 kHz range matters even more. EQ each voice on its own track rather than the whole mix.

Will EQ help me pass ACX?

Indirectly. A high-pass that removes rumble can lower the noise floor reading between sentences, which helps with ACX's noise requirement, but EQ does not set loudness or peaks. Those come from compression, normalisation and limiting at the end of the chain.

Sources and Further Reading

• DPA Microphones, Facts about speech intelligibility: the 2 kHz band as most important for intelligibility, the 1 to 4 kHz range, consonants at 2 to 4 kHz, the 5 percent loss from removing everything below 500 Hz, and the spectral shift when speaking louder.
• Ingo R. Titze, Principles of Voice Production (Prentice Hall, 1994): typical speaking fundamental ranges of about 85 to 180 Hz for adult male voices and 165 to 255 Hz for adult female voices.
• Dennis Bohn, Rane, RaneNote 170: Bandwidth in Octaves Versus Q in Bandpass Filters: the definition of bandwidth at the -3 dB points and the formulas used for the Q graphic.
• Sound On Sound, Vocal de-essing techniques: sibilance concentrated between about 4 and 10 kHz, and the limits of automatic de-essing.
• Sound On Sound, SPL De-Esser Collection review: the male/female switch at 6 and 7 kHz.
• Sound On Sound, Q. Should EQ be used before or after compression?: Mike Senior on why the order matters little for subtle settings.
• Adobe, Filter and equalizer effects in Audition: the Parametric Equalizer, Constant Q, and Q values of 1 to 10 for general equalization.
• Adobe, Amplitude and compression effects in Audition: the DeEsser's threshold, centre frequency, bandwidth, Broadband and Multiband modes, and Output Sibilance Only.
• Audacity, High-Pass Filter and Filter Curve EQ manual pages: menu locations and controls.
• Audacity Forum, Filter Curve 100 Hz rumble fix: the Low rolloff for speech preset in Filter Curve EQ.
• Shure, SM7B user guide: the bass rolloff and presence boost switches, and close-miking for a warmer bass response.