Why wind noise is so hard to fix in Audacity

Wind noise is not one sound. It is usually a low-frequency rumble below 150 Hz, plus intermittent buffeting bursts that can hit 200–500 Hz and higher when the mic capsule overloads. That matters because Audacity's manual tools are great at removing steady low rumble, but they struggle when wind overlaps speech. You can still get a clean result, but you need to know which tool to use and when to stop. If you can re-record with a foam windshield or a dead cat, do that first; it beats every repair tool.

Manual Audacity workflow for wind noise

1. Prep and listen

Import your file. Duplicate the track first: select the track, then Edit > Duplicate (or Ctrl+D). Rename the copy 'working' and leave the original muted. Listen with headphones and mark every windy section. Wind often looks like a dark, wavy cloud below 200 Hz in Spectrogram view.

2. High-Pass Filter

Select all, or just the windy clip. Go to Effect > High-Pass Filter (in newer Audacity: Effect > EQ and Filters > High-Pass Filter). Start at 80 Hz for a deep male voice, 100 Hz for most speech, and 120 Hz for phone or Zoom recordings. Use a 12 dB/octave rolloff. Preview before applying. If the voice sounds thin, lower to 70–80 Hz. If rumble remains, raise to 120 Hz, but listen for lost warmth.

3. Filter Curve EQ

If a boxy 'woof' remains, open Effect > Filter Curve EQ. Make a mild cut of 2–4 dB between 150–300 Hz with a wide curve. Avoid deep, narrow cuts. This reduces wind body without hollowing out the voice.

4. Noise Reduction

Find 0.5–1 second of wind-only audio. Select it and choose Effect > Noise Reduction > Get Noise Profile. Then select the whole track and reopen Noise Reduction. Start with Reduction 6–10 dB, Sensitivity 6.00, and Frequency smoothing 3. Preview. If it sounds metallic or underwater, undo and lower Reduction to 4–6 dB. Applying 6 dB twice is usually cleaner than 12 dB once. Noise Reduction works best on steady hiss or rumble; it tracks intermittent wind poorly.

5. Spectral Delete for isolated gusts

Switch the track to Spectrogram view. Set the window size to 1024 or 2048. Choose the Spectral Selection tool and drag a box around the low-frequency wind blob, usually below 150–200 Hz. Then choose Effect > Spectral Delete. Do this only on gusts between words. If you delete low frequencies under speech, the voice will sound thin and robotic.

6. Gate or mute pauses

If wind appears only in pauses, use Edit > Split or the scissors tool to cut those sections out, or mute them. If your Audacity version has Effect > Noise Gate, start with threshold -35 dB, attack 10 ms, hold 100 ms, decay 200 ms, and floor -80 dB. Raise the threshold if wind leaks; lower it if word tails get chopped.

7. Final check and export

A/B against the original. Listen on phone speakers and headphones. If the voice sounds underwater, undo the Noise Reduction and try a lighter high-pass first. Export WAV for further editing or MP3 at 192 kbps or higher for final delivery.

Manual vs AI voice isolation: the real tradeoff

Manual editing gives you full control, works offline, and costs nothing. But it is slow, and it cannot invent a clean voice if wind is louder than speech. High-pass filtering removes rumble but not midrange wind. Noise Reduction can create metallic artifacts. Spectral editing is precise but tedious. For a 30-minute podcast with intermittent wind, you can spend an hour and still hear gusts.

AI voice isolation uses a neural network trained to separate speech from noise. Instead of guessing a frequency band, it predicts which parts of the spectrogram are voice and which are wind, traffic, fans, or room noise. That is why it handles overlapping wind—where a gust happens during a sentence—much better than a static EQ or noise profile. A free online option is ClearVoiceHQ (clearvoicehq.com): upload your audio or video, run voice isolation, and download the cleaned track. It is especially useful for phone recordings, Zoom calls, gaming clips, and voice memos where you cannot re-record. Then use Audacity for final high-pass, EQ, and loudness. ClearVoice is not a substitute for good mic placement, but it can turn an unusable take into a usable one.

Best hybrid workflow

  1. Back up the original file.
  2. If wind overlaps speech, run AI voice isolation first, such as ClearVoice.
  3. Import the cleaned file into Audacity.
  4. Apply a high-pass filter at 80–100 Hz with a 12 dB/octave rolloff.
  5. Make a light EQ cut at 200–300 Hz if the voice still sounds boxy.
  6. Use Noise Reduction only if steady hiss remains: 4–6 dB, Sensitivity 6, Smoothing 3.
  7. Normalize to -1 dB peak, or use Loudness Normalization: -16 LUFS for podcast, -14 LUFS for YouTube.
  8. Export and listen on two different systems.

Settings cheat sheet

Final word

Start with the least destructive fix: a high-pass filter. If that fails, use Spectral Delete on isolated gusts. If wind overlaps speech and you need it gone fast, run AI voice isolation first, then finish in Audacity. The best result usually comes from a hybrid approach—not from one magic slider. Always process a copy and keep the original.