Editing Podcast Episodes to Remove Unwanted Background Noise

Before You Touch a Single Slider

The temptation when you open a podcast episode is to reach straight for the noise reduction and push it until the hiss disappears. Resist that. The most useful thing you can do first is listen properly. Put your headphones on at a comfortable level — loud enough to hear detail, quiet enough that you're not fooling yourself — and play the whole episode from start to finish with a note file open beside you.

Pay particular attention to the gaps. Noise hides under speech and announces itself in silence. Every time you hear something that shouldn't be there, note the timecode and what it is: "04:12 — hum gets louder when the guest leans forward"; "11:38 — click, probably the arm of the chair." Twenty minutes of this saves hours later, because you'll know exactly what you're treating rather than facing one vague problem.

Steady Hums and Occasional Clicks Are Two Different Problems

A podcast's unwanted noise almost always falls into one of two families, and they need different tools.

Steady noise is continuous and predictable: a laptop fan, a fridge two rooms away, traffic rumble, the hiss of a cheap preamp, or mains hum at 50 Hz with harmonics at 100 Hz and 150 Hz from a poorly earthed cable. Because it barely changes from second to second, software can learn its shape and subtract it, which is exactly what noise reduction is good at.

Occasional noise is a one-off: a keyboard tap, a chair creak, a lip smack, a plosive thump, a phone buzzing on the table. These events are short and unique, so a noise profile cannot predict them. They need manual repair — zooming in, selecting the offending few milliseconds and rebuilding them.

Confusing the two is where home edits go wrong: heavy noise reduction smears a click into a puff of mush, while manual patching can never keep up with a constant hum.

Use Noise Reduction Sparingly — Then Back It Off

Noise reduction samples a stretch of pure noise, builds a profile, then removes matching frequencies throughout the file. The key word is matching: the harder you push it, the more of your voice it takes with it.

• Find at least a second of room tone with no speech — before the first word or after the last breath.
• Take the profile from there, not from a pause containing a breath or a rustle.
• Start with a gentle 6–10 dB of reduction and moderate sensitivity.
• Apply two or three light passes rather than one heavy one.

Then bypass it and listen. Over-cooked noise reduction leaves sustained vowels watery and metallic, puts a faint "swimming" tail on the ends of words, and turns sibilants into a lisp. If you can hear the processing working, you've gone too far.

Often a high-pass filter does more good. Rolling off everything below 80 Hz for a lower voice, or 100 Hz for a higher one, removes rumble, desk thumps and much mains hum without touching intelligibility. For a narrow tonal hum that survives, a notch filter at 50 Hz and its harmonics is cleaner than a broad brush.

Repairing Clicks, Coughs and Stray Sounds

One-off noises are surgical work. Zoom in until you can see individual waveform cycles, then select only the few milliseconds containing the click. Most editors offer a redraw or spectral repair tool that rebuilds the selection from the surrounding audio. If yours doesn't, reduce the gain on the selection rather than deleting it — a quiet click distracts far less than a hole.

Give every cut a short crossfade at each end, typically 5–15 milliseconds. Without one you'll create a tiny click of your own, which is an irritating way to spend an afternoon.

For a cough or a chair scrape over a word, look for the same phrase elsewhere in the take. Guests repeat themselves constantly, and a clean "so what we did next" from two minutes earlier will often paste in invisibly. Before you begin, record thirty seconds of silence in the room and keep it; that room tone is invaluable for patching gaps and covering edit seams.

Check the Result on Speakers and Headphones

Never trust one listening device. Headphones are unmatched for detail — they'll reveal a faint hum, a clumsy edit point, a metallic artefact — but they flatter the low end and can make a noisy file sound cleaner than it is. Speakers in a room show you how the episode behaves in the real world, and a small Bluetooth or laptop speaker is a brutal test for harshness.

• Play the same thirty seconds on headphones, then speakers, then a phone.
• Toggle the noise reduction on and off so you're comparing, not remembering.
• Listen closely to the quiet passages; that's where the treatment shows.
• Check a full paragraph at normal speed, not slowed down.

A Final Pass, and Knowing When to Stop

Once the repairs are done, play the episode end to end without stopping. Pausing every ten seconds breaks your sense of flow, and flow is what listeners actually notice. Keep a raw, untouched copy of the recording archived — storage is cheap, and you'll be grateful for it if an edit goes wrong later.

Export with sensible headroom: peaks no higher than about −1 dB, and a loudness target of roughly −16 LUFS for a stereo podcast. Then listen to the finished file the next morning on whatever device is nearest. Fresh ears catch what tired ones forgive.

And when it sounds good enough, stop. A slightly imperfect episode published this week beats a flawless one that never leaves your hard drive. Background noise only matters when it distracts from what's being said — the aim isn't silence, it's a recording that lets the conversation through.