First, identify the noise: white noise vs hum vs room tone
White noise is broadband hiss—roughly equal energy across frequencies. It sounds like a fan, air conditioner, cheap preamp, or tape hiss. Hum is different: a low 50/60 Hz tone plus harmonics. Room tone is the quiet ambience of a room. Audacity can reduce all three, but the settings differ. If you use Noise Reduction on a hum without a notch or high-pass filter, you may still hear the hum. If you gate room tone too hard, you get unnatural silence. Match the tool to the noise first.
Manual Audacity workflow for white noise
1. Find a clean noise sample
Select 1–2 seconds where only noise is present—before the first word, between sentences, or after the last word. If your recording has no pure noise gap, select the quietest 0.5–1 second you can find. Open Effect > Noise Removal and Repair > Noise Reduction (in older Audacity, Effect > Noise Reduction), then click Get Noise Profile. Audacity uses that selection to learn the hiss fingerprint. A voice-contaminated profile will teach Audacity to remove voice, so be picky.
2. Apply Noise Reduction in small passes
Select the whole track. Open Noise Reduction again. Start with these settings:
- Noise reduction: 6–10 dB for mild hiss; 10–12 dB for heavy hiss. Avoid 20+ dB unless the recording is unusable.
- Sensitivity: 6.00–8.00. Higher values catch more noise but also remove breaths, reverb tails, and consonants.
- Frequency smoothing: 3–6 bands. Use 3 for speech, 6 if the result sounds harsh or metallic.
Click Preview. If the voice sounds watery, robotic, or lispy, cancel and lower the noise reduction by 3–4 dB. Two passes at 5 dB often beat one pass at 12 dB. For a second pass, do not click Get Noise Profile again unless the noise changed; just apply the same profile with gentler settings.
3. High-pass filter the rumble
White noise often comes with low-frequency rumble from traffic, desk bumps, or air handling. Select all and use Effect > EQ and Filters > High-Pass Filter. For most male voices, set frequency to 70–90 Hz and rolloff to 12 dB per octave. For female or higher voices, try 100–120 Hz. This will not erase hiss, but it cleans up the low end and helps Noise Reduction work less aggressively.
4. Gate or manually silence gaps
If hiss is still audible between phrases, use a gate. In current Audacity, look under Effect > Noise Removal and Repair > Noise Gate. Start with a threshold just above the noise floor—often -45 to -35 dB for spoken voice. Set level reduction to -12 dB, not -80 dB, so the silence does not sound chopped. Use attack 10–20 ms, release 100–200 ms, and hold 50–100 ms. Preview and adjust. For short recordings, manual silencing can sound better: select a noise-only gap, press Ctrl+L to silence it, and repeat. It is tedious but artifact-free.
5. Use spectral repair only for tones, not broadband hiss
Audacity's spectral tools are useful for a whistle, coil whine, or a constant beep. They are not the best fix for white noise because white noise is spread across all frequencies. If you see a bright horizontal line in Spectrogram view, select that frequency range and use Effect > Spectral Delete (or a similar spectral repair tool). For general hiss, stick with Noise Reduction.
6. Finish with EQ and compression
After cleanup, the voice may sound dull. Add a gentle high-shelf boost around 2–5 kHz if needed. Use Filter Curve EQ for broad tone shaping, not for noise removal. If you compress after Noise Reduction, the compressor will raise any remaining noise, so apply a light downward expander or gate before compression. A simple chain works: high-pass, Noise Reduction, gate, EQ, compressor, limiter.
Common Audacity mistakes and tradeoffs
- Over-processing: too much Noise Reduction makes voices sound underwater. Less is almost always better.
- Bad noise profile: if the profile includes breath or voice, Audacity removes those frequencies from the whole track.
- Aggressive gating: a gate set at -25 dB may cut off word tails and create pumping.
- Working on MP3: compression artifacts limit how much you can clean. Use the original WAV if possible.
- Ignoring room treatment: no plugin fully fixes a noisy room. A blanket behind the mic or a closer mic position helps more than extreme settings.
AI voice isolation vs manual Audacity editing
Manual Audacity editing gives you control. You decide how much noise to remove, which frequencies to cut, and where to silence gaps. It works well for steady, low-level white noise and for recordings where the voice is already clear. The tradeoff is time and skill: a 30-minute podcast can take an hour of tweaking if the noise is uneven.
AI voice isolation takes a different approach. Instead of learning a noise profile, it separates speech from everything else using a trained model. That can remove white noise, hum, fan noise, keyboard clicks, and even some background chatter in one pass. It is often faster for Zoom calls, phone memos, gaming clips, and interviews recorded in imperfect rooms. The tradeoff is less control. AI can introduce artifacts, thin out the voice, or struggle with overlapping speakers and music. For a quick first pass, a free tool like ClearVoice can isolate the voice, then you can finish in Audacity. Always compare the AI result to your manual version; sometimes a light Audacity pass sounds more natural.
Hybrid workflow: AI first, Audacity second
- Upload your original recording to ClearVoiceHQ (clearvoicehq.com) and download the isolated voice track as WAV.
- Import the WAV into Audacity.
- Apply a high-pass filter at 80 Hz, 12 dB per octave.
- If any hiss remains, run Noise Reduction at only 3–6 dB with sensitivity 6 and smoothing 3.
- Add EQ and light compression, then export.
This hybrid approach often avoids the watery artifacts of heavy manual Noise Reduction while keeping the final polish in your hands.
Fastest manual Audacity fix for white noise
- Open the file and find a 1–2 second noise-only section.
- Select it, then Effect > Noise Removal and Repair > Noise Reduction > Get Noise Profile.
- Select all. Open Noise Reduction. Set noise reduction to 8 dB, sensitivity 6.00, smoothing 3. Preview and apply.
- Apply High-Pass Filter at 80 Hz, 12 dB per octave.
- Open Noise Gate. Set threshold around -40 dB, level reduction -12 dB, attack 10 ms, release 150 ms. Preview and apply.
- Listen. If artifacts appear, undo and try 5 dB Noise Reduction instead.
- Export as WAV for editing or 192–320 kbps MP3 for sharing.
Bottom line
To remove white noise from audio in Audacity, start with a clean noise profile, apply Noise Reduction in gentle passes, high-pass the rumble, and gate only the quiet gaps. If the noise is complex or you have many files, AI voice isolation can get you 80–90% of the way in one click. Use manual Audacity for control and AI for speed, then trust your ears—if the cleanup sounds unnatural, you have gone too far.