guides

Podcast Noise Reduction Techniques: Remove Background Noise Without Artifacts

PodRewind Team
8 min read
sound wave visualization showing noise filtering process
Photo via Unsplash

TL;DR: Effective noise reduction matches the technique to the specific noise type. Broadband noise responds to spectral subtraction, hums need notch filters, reverb requires specialized de-reverb tools. Always capture a noise profile from silence, use conservative settings, and accept that slight background noise sounds better than over-processed, artifact-heavy audio.


Table of Contents


Understanding Podcast Noise Types

Different noises require different treatments. Identifying the noise type determines which tool to use.

Here's the thing: a single "noise reduction" setting rarely fixes all problems. Professional results come from addressing each noise type with the appropriate technique.

Broadband Noise

What it sounds like: Constant hiss across all frequencies, like white noise or static.

Common sources:

  • Preamp self-noise (especially budget equipment)
  • Air conditioning systems
  • Computer fans
  • Electrical interference

Treatment: Spectral subtraction or noise profiling tools

Tonal Noise (Hums and Buzzes)

What it sounds like: Consistent pitch, often a low hum or high-pitched whine.

Common sources:

  • Electrical ground loops (60Hz in US, 50Hz elsewhere)
  • Fluorescent lighting
  • Electronics interference
  • Refrigerators and appliances

Treatment: Notch filters or de-hum tools targeting specific frequencies

Room Reverb

What it sounds like: Voices sound distant, hollow, or echoey. Speech has a "tail" that lingers.

Common sources:

  • Recording in untreated rooms
  • Hard surfaces (walls, floors, windows)
  • Large spaces with parallel walls

Treatment: De-reverb processing (limited effectiveness) or prevention

Transient Noise

What it sounds like: Sudden sounds that interrupt recording.

Common sources:

  • Mouth clicks and lip smacks
  • Keyboard typing
  • Page turns
  • Handling noise from microphones

Treatment: Manual editing, de-clicking tools, or spectral repair

Intermittent Background Noise

What it sounds like: Noise that comes and goes throughout recording.

Common sources:

  • Traffic passing by
  • HVAC cycling on and off
  • People in adjacent rooms
  • Outdoor sounds through windows

Treatment: Manual editing during quiet sections, noise gates with care, or spectral repair


Capturing an Effective Noise Profile

Most noise reduction tools require a "noise profile"—a sample of the unwanted noise without any speech.

Finding Clean Noise Samples

Ideal noise sample characteristics:

  • 2-5 seconds of pure background noise
  • Recorded during the same session (noise changes over time)
  • Representative of the typical noise floor
  • No speech, breaths, or other sounds

Where to find noise samples:

  • Pauses between speakers
  • Beginning or end of recording before/after conversation
  • Moments where host asked guest to pause

Recording Room Tone Intentionally

Best practice: record 30-60 seconds of silence at the beginning or end of every session.

Why room tone matters:

  • Ensures you always have a clean noise profile
  • Captures the exact noise conditions of that recording
  • Useful for patching edits and filling gaps

Ask everyone to remain silent while you capture room tone. Any movement or sound contaminates the sample.

Profile Selection Mistakes

Selecting speech as part of the profile:

If your noise profile contains any voice, the noise reduction will treat voice frequencies as noise. Result: hollow, distorted, or artifact-heavy speech.

Selecting atypical noise:

A brief loud noise (truck passing, cough) creates an inaccurate profile. Find a section that represents the consistent background, not momentary sounds.

Profile too short:

Very short profiles (under one second) may not capture the full frequency range of the noise. Use at least 2 seconds.


Spectral Noise Reduction

Spectral noise reduction analyzes the frequency content of your noise profile and subtracts it from the entire recording.

How It Works

  1. The tool analyzes your noise sample to create a frequency "fingerprint"
  2. During playback, it identifies those frequencies in the audio
  3. It reduces or removes those specific frequencies
  4. Speech frequencies remain (mostly) intact

Standard Settings Explained

Noise Reduction (dB):

How much to reduce the identified noise. Higher values remove more noise but risk artifacts.

  • Conservative: 6-12dB
  • Moderate: 12-18dB
  • Aggressive: 18-30dB (high artifact risk)

Sensitivity/Threshold:

How aggressively the tool identifies what counts as "noise."

  • Low sensitivity: Only the most obvious noise gets treated
  • High sensitivity: More frequencies identified as noise (including some speech)

Frequency Smoothing:

Prevents the reduction from sounding choppy by smoothing between frequency bands.

For podcast speech with typical background noise:

SettingStarting Value
Noise Reduction10dB
Sensitivity4-6
Frequency Smoothing3-5 bands

Adjustment approach:

  1. Start with conservative settings
  2. Apply and listen to a variety of sections
  3. If noise still obvious, increase reduction slightly
  4. If artifacts appear, reduce settings or increase smoothing
  5. Find the balance point between noise and naturalness

Common Software-Specific Tools

Audacity: Effect > Noise Reduction Adobe Audition: Effects > Noise Reduction/Restoration > Noise Reduction iZotope RX: Voice De-noise or Spectral De-noise Descript: Background noise removal (automatic)


Dealing with Specific Noise Problems

HVAC and Ventilation Hum

Air conditioning systems produce consistent low-frequency noise, often with tonal components.

Approach:

  1. Use spectral noise reduction for broadband component
  2. Add a high-pass filter at 80-100Hz to remove rumble
  3. If tonal hum remains, use notch filter at the specific frequency

Prevention: Turn off HVAC during recording or record in rooms with quieter systems.

Electrical Hum (Ground Loops)

60Hz hum (US) or 50Hz hum (elsewhere) from electrical interference.

Identification: Distinctive constant buzz at a single pitch.

Treatment:

  1. Apply notch filter at fundamental frequency (60Hz or 50Hz)
  2. Add notch filters at harmonics (120Hz, 180Hz, etc.)
  3. Use dedicated de-hum tools that target all harmonics automatically

Prevention: Fix ground loops in your recording chain, use balanced cables, separate audio cables from power cables.

Room Echo and Reverb

Reverb is the hardest noise problem to fix in post-production.

Why it's difficult:

  • Reverb frequencies overlap with speech frequencies
  • Removing reverb also removes parts of the voice
  • De-reverb processing often sounds unnatural

Treatment approach:

  1. Use de-reverb tools conservatively (many DAWs now include these)
  2. Accept that results will be limited
  3. Heavy reverb often can't be fully fixed

Prevention: Acoustic treatment before recording is far more effective than post-production repair. Even temporary solutions (blankets, closet recording) help significantly. Once you have clean audio, a consistent editing workflow ensures quality across every episode.

Mouth Clicks and Lip Smacks

Quick transient sounds from dry mouth or natural speech patterns.

Treatment options:

  1. De-click processors: Automatic detection and reduction
  2. Spectral repair: Manual removal in spectral editing view
  3. Volume reduction: Lower the volume of individual clicks manually

Settings for de-clickers:

  • Use low sensitivity to avoid affecting consonants
  • Process in multiple light passes rather than one aggressive pass

Traffic and Outdoor Noise

Variable noise that changes throughout recording.

Treatment:

  1. Standard noise reduction helps with consistent traffic rumble
  2. High-pass filter removes low-frequency rumble
  3. Manual editing for obvious individual events
  4. Noise gates (used carefully) during pauses

Challenge: Variable noise doesn't have a single profile—treatment effectiveness varies throughout the episode.


When Noise Reduction Causes More Harm

Aggressive noise reduction creates problems worse than the original noise.

Recognizing Artifacts

Warbling or "underwater" sound:

Over-processed audio sounds like speaking through a pipe or underwater. Caused by excessive reduction removing parts of speech.

Musical artifacts (birdies):

Random tonal sounds that weren't in the original. Caused by noise reduction creating new frequency content.

Pumping or breathing:

The noise floor rises and falls audibly around speech. Caused by noise gates or extreme threshold settings.

Metallic or robotic quality:

Voices lose natural warmth and sound artificial. Caused by removing too much frequency content.

When to Accept Background Noise

Sometimes the best edit is minimal processing.

Accept noise when:

  • Reduction causes audible artifacts
  • Noise is consistent and listeners adapt quickly
  • Heavy processing makes speech hard to understand
  • The noise is actually quiet and you're being too perfectionist

Listeners tolerate low-level consistent noise much better than obvious processing artifacts.

The 3dB Rule

If reducing noise by an additional 3dB creates noticeable artifacts, stop where you are. The last 3dB of reduction rarely provides listener benefit worth the artifact cost.


Multi-Stage Noise Reduction Workflow

For recordings with multiple noise problems, address each issue separately in the optimal order.

  1. Remove DC offset (if present)
  2. High-pass filter (removes rumble below speech frequencies)
  3. Hum removal (notch filters for electrical interference)
  4. Spectral noise reduction (broadband noise)
  5. De-reverb (if needed and effective)
  6. De-click/de-ess (transient problems)
  7. EQ and compression (normal processing chain)

Why Order Matters

High-pass before spectral reduction:

Removes low-frequency noise that spectral tools struggle with, letting them focus on relevant frequencies.

Hum removal before spectral:

Spectral reduction works better when consistent tonal noise is already removed.

De-click after spectral:

Spectral noise reduction can create or emphasize clicks; de-clicking afterward catches these.

Processing in Passes

Rather than one aggressive treatment, use multiple lighter passes.

Example workflow:

  1. First pass: 8dB reduction, moderate sensitivity
  2. Listen and evaluate
  3. Second pass: 4-6dB reduction if needed
  4. Listen again—often two passes sounds better than one 12-14dB pass

FAQ

How much noise reduction is too much?

When you hear artifacts—warbling, underwater sounds, or metallic quality—you've gone too far. Most podcast recordings sound best with 6-12dB of reduction. Beyond 15-18dB, artifacts become likely. If noise remains prominent after 15dB reduction, the recording may need re-recording or acceptance of the noise level.

Can I completely remove background noise from a podcast?

Complete removal is rarely possible without damaging voice quality. The goal is reduction to a level where noise doesn't distract listeners, not elimination. Aim for noise that disappears when speech is present and is only barely audible during pauses. Listeners adapt to consistent low-level noise quickly.

Why does my noise reduction make voices sound hollow?

The noise profile likely contained some speech frequencies, or settings are too aggressive. Recapture your noise profile from a section with pure silence. Reduce the noise reduction amount and sensitivity. Add frequency smoothing to prevent choppy processing. Some hollowness may require accepting less noise reduction.

Should I apply noise reduction before or after EQ?

Apply noise reduction before EQ and compression. Noise reduction works best on unprocessed audio, and EQ changes would alter the noise profile match. The standard order is: noise reduction, then EQ, then compression, then limiting. This produces the cleanest results.

How do I remove noise from a recording without a clean noise sample?

Some tools can estimate noise from the audio itself, though results vary. Audacity's "Get Noise Profile" can sometimes work with very short pauses. iZotope RX has adaptive modes that don't require profiles. For best results with difficult recordings, use spectral editing to manually paint out noise in the frequency display.



Ready to Clean Up Your Audio?

Noise reduction transforms distracting recordings into listenable episodes. Match your technique to your noise type, use conservative settings, and prioritize natural-sounding speech over silent backgrounds.

Clean recordings deserve permanent preservation. Transcription transforms your noise-reduced episodes into searchable archives where every clear word becomes findable.

Try PodRewind free and make your polished audio part of a searchable archive.

noise-reduction
audio-quality
editing
post-production

Ready to Get Started?

Search your podcast transcripts, chat with your archive, and turn episodes into content. Start for free today.

Try PodRewind free