How to Remove Podcast Background Noise Without Artifacts

You finish a great conversation, open the file, and hear it under every pause: a fridge hum, a laptop fan, traffic two streets away. Podcast background noise is the most common complaint I hear from new hosts, and the usual fix, dragging a noise reduction slider to the right, often makes the episode sound worse than the noise did.

This guide covers how to identify the kind of noise you have, how to remove it in the right order, and where to stop so your voice still sounds like a person in a room rather than a voice coming through a tin can.

Know which podcast background noise you are dealing with

Different noises need different tools. Before you touch a plugin, solo a few seconds of room tone, the gap between sentences, and listen on headphones at a decent level.

Broadband hiss. A steady "shhh" across the whole spectrum. Usually comes from cheap preamps, gain set too high, or computer fans.
Tonal hum. A low, steady note at 50 Hz or 60 Hz depending on your country, often with harmonics at 100, 150 or 120, 180 Hz. Caused by ground loops, lighting, or power cables near audio cables.
Intermittent noise. Dogs, doors, car horns, keyboard clicks. These come and go and do not respond well to standard noise reduction.
Room reflections. Not noise exactly, but a hollow, echoey quality. Noise reduction will not fix this, and trying usually makes it worse. That is a room problem, covered in how to tame a boomy room for cleaner podcast audio.

Most recordings have two of these at once. Write down which ones you hear, because the order you treat them in matters.

Fix it at the source first

Ten minutes of setup beats an hour of repair. Before your next session, try these:

Turn off the fridge, air conditioner and any fans for the length of the recording. Put a note on the fridge door so you remember to turn it back on.
Move the microphone closer. Going from 30 cm to 10 cm from your mouth raises your voice by roughly 9 to 10 dB relative to the room, which means you can lower gain and the noise floor drops with it.
Use a dynamic microphone if your room is noisy. Dynamics pick up far less from the sides and back than condensers.
Plug your interface and computer into the same power strip to reduce ground loop hum.
Record 10 seconds of silence at the start of every session. You will need this room tone sample later.

Cheap fixes like these are covered in more depth in the cheapest upgrades that actually improve podcast audio, and most cost nothing.

A cleanup chain that works in any editor

In Audacity, Reaper, Adobe Audition or Descript, the order of operations is the same. Run these on each speaker's track separately, not on the final mix.

Step 1: High-pass filter

Set a high-pass filter at 70 to 80 Hz for most voices, or 100 Hz for higher voices. This removes rumble from traffic, footsteps and desk bumps with no audible effect on speech. It is the single safest cleanup move you can make.

Step 2: Notch out hum

If you hear tonal hum, use a narrow EQ notch or a dedicated de-hum tool at 50 or 60 Hz and the first two or three harmonics. Keep the notch narrow, a Q of 10 or higher, so you only remove the hum and not the warmth of the voice.

Step 3: Spectral noise reduction

Select your room tone sample, capture it as the noise profile, then apply reduction to the whole track. Start gentle: 6 to 10 dB of reduction in Audacity, or around 40 to 50 percent in tools that use a percentage. Preview, then raise it slowly until the hiss is quieter but not gone.

Step 4: Gate or expander, lightly

A gentle expander with 4 to 6 dB of range can push down the noise in pauses without chopping off word endings. Avoid a hard gate that drops to silence. Listeners notice when the room vanishes between sentences.

Step 5: Handle intermittent sounds by hand

For a single dog bark or door slam, use a spectral editor to paint it out, or simply cut it if it falls between sentences. Automated tools struggle with these because they do not repeat.

Why you should leave some noise in

The standard advice is to remove as much background noise as possible. I disagree, and the reason is artifacts. Every noise reduction algorithm works by guessing which parts of the signal are noise. Push it too far and it starts guessing wrong, which produces a watery, swirling sound sometimes called "musical noise," along with dulled consonants and a flat, lifeless tone.

Listeners tolerate a low, steady hiss easily. Their ears tune it out within a few seconds. They do not tolerate a voice that sounds processed, because it reads as cheap and it is tiring over a 45-minute episode. A noise floor around minus 60 dBFS in the pauses is perfectly acceptable for spoken audio. You do not need minus 80.

My rule: reduce until the noise stops being the first thing you notice, then back off by 2 or 3 dB. If you cannot hear the difference with the processing bypassed during speech, you have done it right.

About AI noise removal tools

Tools like Adobe Podcast Enhance, iZotope RX Voice De-noise and the Studio Sound feature in Descript can rescue recordings that standard tools cannot. They work by rebuilding the voice rather than subtracting noise, which is why they handle intermittent sounds better.

The trade-off is character. At full strength, these tools can make every voice sound similar, thin out laughter, and smear crosstalk when two people speak at once. Most offer a strength or mix control. Somewhere between 40 and 70 percent usually keeps the voice natural while removing most of the problem. Always compare against the original on headphones before committing.

Check the result before you publish

Listen to three spots: the quietest pause, the loudest laugh, and a stretch where two people talk over each other. If the pause sounds natural, the laugh still sounds like a laugh, and the crosstalk is not garbled, you are done. If one speaker was cleaned more than another, the shift between them will be obvious, so match the processing across tracks. Recording guests in different rooms makes this harder, and recording a remote podcast guest without losing quality covers how to keep their tracks closer to yours from the start.

Frequently asked questions

How do I remove background noise from a podcast for free?

Audacity has a free Noise Reduction effect that works well for steady hiss and hum. Select a few seconds of room tone, click Get Noise Profile, then apply the effect to the whole track at 6 to 10 dB. Add a high-pass filter at 80 Hz first for best results.

Why does my podcast sound robotic after noise reduction?

You applied too much reduction, and the algorithm started removing parts of your voice along with the noise. Undo and try again at half the strength. It is better to leave a little steady hiss than to create artifacts that tire listeners.

Should I apply noise reduction before or after compression?

Before. Compression raises quiet parts of the signal, including the noise floor, so compressing first makes noise louder and harder to remove cleanly. Clean each track, then compress, then set final loudness.