The short answer
Every sound is a blend of frequencies, from the low rumble of a truck to the bright sparkle of a cymbal, all stacked together in one signal. An audio filter is a tool that treats those frequencies differently: it lets some pass through untouched and turns others down. The range it keeps is called the passband, and the range it removes is the stopband. Nothing else about the sound changes, only how much of each pitch survives.
The two filters you will run into constantly are named for the part of the spectrum they let through. A high-pass filter passes the highs and cuts the lows, so it strips away rumble. A low-pass filter passes the lows and cuts the highs, so it softens hiss and harshness. That is the whole idea, and almost everything else is a detail about how sharply and where the cut happens.
High-pass and low-pass, plainly
The names describe what gets through, not what gets removed, which trips people up at first. Read them literally. A high-pass lets the high frequencies pass, so everything below a chosen point is turned down. A low-pass lets the low frequencies pass, so everything above a chosen point is turned down. Put one of each on the same signal and they frame a band in the middle: the lows are gone, the highs are gone, and what remains is the useful range between them.
This is exactly how you clean up a voice. Human speech lives mostly in the middle of the spectrum, while the trouble tends to gather at the edges. Desk thumps, footsteps, and air-conditioning drone sit down at the bottom, below anything your voice needs. Tape hiss, fan whine, and piercing sibilance sit up at the top. A high-pass clears the bottom edge and a low-pass eases the top edge, leaving the part that carries your words.
You do not always use both at once. A high-pass alone is the single most common filter in audio, because low rumble is nearly always unwanted and rarely part of the sound you care about. A low-pass gets reached for less often and more gently, since the highs carry clarity and cutting too much of them makes a voice sound dull or muffled.
Cutoff and slope
Two settings define any of these filters. The cutoff frequency is the point where the filter starts working, measured in hertz. Set a high-pass cutoff to 100 Hz and it leaves everything above 100 Hz alone while turning down everything below it. By convention the cutoff is the frequency where the signal has dropped by 3 decibels, which marks the edge of the passband rather than a hard wall.
The word slope describes how quickly the filter cuts once it starts, measured in decibels per octave (an octave is a doubling of frequency). A gentle slope, like 12 dB per octave, eases the sound down over a wide range and sounds natural. A steep slope removes the stopband more aggressively but can sound abrupt. A filter is never a cliff: it is a ramp, and the slope is how steep that ramp is. The related idea of Q, or resonance, controls how tight the corner is right at the cutoff, but for cleaning a voice you want it gentle and unresonant.
- Cutoff. Where the filter acts. Move it up and a high-pass removes more; move it down and it removes less. The art is placing it just outside the range you want to protect.
- Slope. How hard it acts past the cutoff. Gentle slopes are forgiving and transparent; steep slopes are surgical but easier to overdo.
Filters, EQ, and where they overlap
A filter and an equalizer are close relatives. An EQ shapes the balance of the spectrum, and it can boost as well as cut, lifting presence here or dipping muddiness there. A high-pass and a low-pass are the pure-cut members of that same family: they only ever take frequencies away, never add them. In fact most EQs include a high-pass and a low-pass as their first and last stages, precisely because clearing the extremes is step one of almost every clean-up.
Because a filter only removes, it is a gentle, low-risk tool. That makes it a natural first move in a processing chain, ahead of anything that reacts to level. Removing rumble before it reaches a compressor or a noise suppressor means those tools are not triggered by energy you were going to throw away anyway. All of this runs as DSP, the real-time math that processes your audio as it flows through, so the filtering happens live with no bounce or render step.
Filters in patchd
In patchd, a Windows virtual audio mixer, the EQ on any channel includes both a high-pass and a low-pass filter, so you can trim the low and high edges of a source without touching the middle. A high-pass also pairs naturally with noise suppression: clearing low rumble first leaves the suppressor less to work around.
The practical recipe is simple. Put a gentle high-pass under a voice to remove desk bumps and HVAC drone, setting the cutoff just below where the voice sits so you never thin it out. If a source has a steady top-end hiss, a low-pass can tame it, but reach for it lightly, since the highs are where clarity lives. The guiding rule with filters is restraint: a little removes the problem, and too much removes the life. All of it runs in the real-time engine, which adds only single-digit milliseconds inside the bus at a small ASIO buffer, so the cleanup happens live while you talk.
The takeaway
An audio filter passes some frequencies and blocks others. A high-pass keeps the highs and clears the low rumble; a low-pass keeps the lows and softens the high hiss; together they frame the useful band in the middle. Two settings do the work: the cutoff, which places the filter, and the slope, which sets how hard it cuts. Used gently, filters are the cleanest, safest way to take a problem out of a signal without changing the part you want.
patchd is pre-launch, with a per-channel EQ and noise suppression, high-pass and low-pass included, on every tier. If you want a Windows mixer that keeps your audio clean without stacking up delay, join the waitlist and get notified when it ships. The paid Studio tier is $39.99 per year, and the real-time engine is identical on both tiers.