Tutorials Audio Engineering

How to Remove Background Noise from Audio

Beginner · ~30 min

Overview

No noise removal tool can fully recover a recording made in a loud room, prevention is always better than cure. But for steady background noise like air conditioning, fan hum, and computer noise, modern tools (including free ones) can remove it cleanly with minimal artefacts if used correctly.

What You Need

Your noisy recording (WAV preferred)
Audacity (free) for basic noise reduction
Optional: NVIDIA RTX Voice, Krisp, or Adobe Podcast Enhance (free AI tools)
Optional: iZotope RX Elements (~$99) for professional results

Steps

Identify the type of noise

Different noise types require different tools. Steady broadband noise (fan hum, air conditioning, tape hiss): Audacity noise reduction works well. Intermittent noise (traffic, keyboard clicks, dog barking): needs a noise gate or manual editing. Variable broadband noise (crowd noise, wind): requires AI tools or iZotope RX. Correct diagnosis prevents wasted time on the wrong approach.

Method 1: Audacity noise reduction (free)

Select a 1 to 2 second section of pure noise (no speech). Go to Effect → Noise Reduction → Get Noise Profile. Select your entire recording (Ctrl+A). Return to Effect → Noise Reduction and set: Noise Reduction 12dB, Sensitivity 6, Frequency Smoothing 3. Click OK. The critical setting is keeping Noise Reduction at 12dB or below, higher values remove more noise but introduce metallic "underwater" artefacts that are worse than the original noise.

Method 2: AI noise removal tools (free)

Adobe Podcast Enhance Speech (free at podcast.adobe.com/enhance) uploads your file and returns a cleaned version using AI. The results are often better than Audacity for mixed noise environments. NVIDIA RTX Voice (free, requires RTX GPU) runs in real time during recording. Krisp has a free tier with monthly minutes. These tools are dramatically easier than manual processing for challenging recordings.

Method 3: iZotope RX Voice De-noise (paid)

iZotope RX is the industry standard for audio repair. Load your file in RX Standalone, apply the Voice De-noise module in Adaptive mode. It continuously learns the noise floor and removes it even as it changes. The result is transparent on steady noise and handles variable noise better than any other tool at the price point. RX Elements (~$99) includes Voice De-noise.

Check for artefacts and adjust

After processing, listen critically on headphones to the transitions between silence and speech. Listen for: metallic resonance on vowels ("underwater" sound), warbling on sustained notes, musical noise (tonal artefacts that weren't there before). If you hear any of these, reduce the noise reduction amount. A slightly noisy but natural recording is always preferable to an artefact-filled "clean" one.

Pro Tips

A noise gate (silences audio below a threshold) combined with light noise reduction is often more natural than heavy noise reduction alone.
Always work on a copy of the original file, noise reduction is destructive and can't be undone without re-processing.
Use our Noise Floor Calculator to understand how SNR affects your recording quality before you reach for noise removal.

Knowledge Base

What You'll Learn

Remove background noise from recordings. Learn when noise reduction works and when it creates artifacts. Master the balance between clean audio and natural-sounding speech/voiceover.

When Noise Reduction Works

Steady Noise: AC hum, fan noise, traffic. Highly removable.
Electrical Hum: 50/60 Hz hum from power. Nearly 100% removable.
Room Tone: Constant background ambience. Partially removable.

When It Struggles

Intermittent Noise: Dog barking, phone ringing, sirens. Can't be removed without distorting audio.
Noise Similar to Voice: Wind hiss, paper rustling. Removal degrades speech clarity.

Noise Reduction Workflow (Audacity)

1) Select 1-2 seconds of pure noise (no speech). 2) Effect → Noise Reduction → Get Noise Profile. 3) Select all audio. 4) Effect → Noise Reduction → Apply (strength 12dB). 5) Listen and adjust (higher = more aggressive, more artifacts).

Best Practices

Use subtle settings first (6-8 dB reduction). If noise remains, apply again.
Don't chase perfect silence. Some noise is acceptable if speech is clear.
Check result on phone speakers, artifacts are often more noticeable there.
Recording in a quiet space first is always better than fixing in post.

Where This Fits

This guide covers one specific part of audio recording. The wider picture, how microphone distance, room acoustics, polar patterns, and gain staging decide whether a recording is usable before any processing, is in Audio Recording Fundamentals: The Complete Guide, which frames the discipline as a whole and links out to the detailed guides underneath it, including this one. If you are starting from scratch rather than solving a specific problem, read that first and come back here.

FAQ

Q: What if noise reduction makes audio sound thin or robotic?
A: You over-applied it. Undo and use lower strength (6-8 dB instead of 12+ dB).