Audacity equalizer: how to use Filter Curve EQ on a voice

If you are following a tutorial that says Effect then Equalization and there is no such item, you are not doing it wrong. Audacity split that single effect into two, and both now live under Effect, EQ and Filters.

Filter Curve EQ is the draw-your-own version: you click on a line to add control points and drag them to shape the response. It is the direct replacement for what the old Equalization effect did in its curve mode.

Graphic EQ gives you a row of sliders at fixed frequencies, like a hi-fi equaliser. It is less precise and considerably easier to use, and for most voice work it is entirely sufficient.

Both carry the same preset list, so the saved curves people share still work in either.

The single most useful EQ move on a spoken recording is a high-pass filter: roll off everything below about 80 Hz. There is almost nothing in a human voice down there, and there is a great deal of rumble, desk vibration, traffic and air conditioning.

In Filter Curve EQ this is a steep downward slope at the left edge of the graph. Audacity also ships a High Pass Filter as its own effect under the same menu, which is simpler if that is all you want.

For a deep male voice, set the cut closer to 70 Hz. For most other voices 80 to 100 Hz is safe. You will not miss anything and the recording immediately sounds cleaner, because you have removed energy that was doing nothing but consuming headroom.

Below 80 Hz: rumble and handling noise. Cut it.

100 to 250 Hz: body and warmth, and where boominess lives if you are close to the microphone. A gentle cut here fixes the sound of someone speaking into a large diaphragm from two inches away.

250 to 500 Hz: the muddy range. If a recording sounds thick or congested, a small cut around 300 Hz usually clears it. Resist boosting here.

2 to 5 kHz: intelligibility. This is where consonants live and where a small boost makes speech easier to follow. It is also where listening fatigue comes from, so a couple of decibels is plenty.

5 to 8 kHz: sibilance, the harsh edge on s and t sounds. If those are painful, cut narrowly here rather than reaching for the whole top end. Above 10 kHz: air and openness, where a gentle lift can make a recording feel less closed in.

The working rule for EQ is that cuts should be narrow and surgical, while boosts should be wide and gentle. A narrow cut removes a specific problem. A narrow boost draws attention to itself and sounds artificial.

Keep the adjustments small. Anything beyond about 6 dB in either direction on a voice is usually solving the wrong problem, and the real fix is microphone position or the room.

Use Preview constantly and toggle it on and off. EQ changes are easy to talk yourself into, and the only honest test is whether the processed version sounds better than the original when you switch quickly between them.

Judge on speakers or headphones you know well, and at a moderate volume. Loud always sounds better, which is why it is a poor way to compare two versions.

A sensible order for spoken audio is: noise reduction, then EQ, then compression, then loudness normalisation. Each step assumes the one before it has already happened.

EQ before compression matters because the compressor reacts to what it is given. Feed it a recording full of low-frequency rumble and the compressor turns the whole voice down every time a lorry goes past outside.

Loudness normalisation goes last, because everything before it changes the level.

If you would rather not build that chain by hand each time, Hilite applies the equivalent processing automatically when it enhances a recording. We make it, so weigh that as you like. Audacity does all of it for free and gives you more control if you want it.