A lecture recording is a bad place to discover background noise. You cannot ask the professor to repeat 45 minutes, and you usually cannot re-record. Most lecture noise—HVAC hum, laptop fan, air conditioner, keyboard clatter, traffic, room echo—can be reduced enough to make speech clear. Work in the right order and stop before the voice sounds underwater.

This guide is for a Zoom local recording, phone voice memo, webinar capture, interview file, podcast clip, or gaming clip with a noisy mic. You do not need a studio. You need the original file, a noise profile, and a conservative hand.

Know what you are removing

Before you edit: source fixes

  1. Use the cleanest source. In Zoom, check whether you recorded separate audio files for each participant. If yes, use your mic track, not the mix. In OBS, record mic to its own track. From a phone, export WAV or M4A without re-compressing it.
  2. If your editor allows it, high-pass at 80–100 Hz. That removes rumble without touching most speech.
  3. For future recordings: mic 4–6 inches from your mouth, slightly off-axis. Peaks around -12 to -6 dB. Turn off fans, use airplane mode, and record 5 seconds of room tone before speaking.

Tradeoff: a live noise gate can make a lecture sound choppy and cut quiet words. If unsure, record raw and clean later.

Step-by-step cleanup in a free editor

These settings work in Audacity. The same ideas apply in Adobe Audition, Reaper, Descript, or Premiere Pro.

1. Import and duplicate

Drag the file into your editor. Save a copy. If the recording is stereo but the speaker is mono, split to mono so you do not process the same noise twice.

2. High-pass filter

Apply a high-pass filter at 80 Hz for deeper voices or 100 Hz for higher voices. Use a 12 dB per octave slope. Do not go to 150 Hz unless the recording is very rumbly; that thins the voice.

3. Fix obvious interruptions manually

Before noise reduction, cut or mute coughs, clicks, and notifications. In Audacity, switch to Spectrogram view, select the noisy blob, and use Repair or Spectral Delete for short clicks. For long gaps, use Truncate Silence with threshold around -40 dB, duration 0.5 seconds, and truncate to 0.2 seconds. Do not cut natural pauses; lectures need breaths.

4. Learn the noise profile

Find 0.5 to 1 second of room tone—no speech, no breathing, no keyboard. Select only that. In Audacity, go to Effect > Noise Removal and Repair > Noise Reduction, then click Get Noise Profile.

5. Reduce noise gently

Select the whole track. Start with Reduction at 6–9 dB, Sensitivity at 6, and Frequency Smoothing at 3. Preview. If noise remains, apply a second pass of 3–6 dB rather than one pass of 18 dB. Two light passes usually sound more natural than one heavy pass. In Adobe Audition, use Noise Reduction with Reduce 6–10 dB, FFT 4096, Smooth 3–6, and Precision Factor 5–7. In iZotope RX, use Spectral De-noise and keep reduction around 6–10 dB.

Tradeoff: too much reduction creates watery, metallic speech, lost consonants, and a lisp on s sounds. If you hear that, undo and use less.

6. De-reverb, gate, and EQ only if needed

If the room echoes, use light de-reverb at 20–40 percent. If you must gate, set threshold around -35 to -45 dB, attack 5–10 ms, release 100–200 ms. That cuts room tone between phrases but can chop quiet speech. For EQ, cut 200–400 Hz by 2–3 dB if the voice is boomy. Add 1–2 dB around 2–5 kHz if words are dull. Use a de-esser if sibilance gets harsh.

7. Normalize and export

Normalize peaks to -1 dB, or target -16 LUFS for online playback. Export WAV at 48 kHz and 24-bit for video editing. For voice-only files, MP3 at 192 kbps or higher is fine. Compare processed and original with bypass to make sure you improved clarity, not just lowered noise.

If you do not want to learn an editor

For a fast first pass, a free tool like ClearVoiceHQ (clearvoicehq.com) can remove steady fan hum, air conditioner rumble, and background hiss from a lecture file without installing anything. Upload the recording, choose a light or medium noise-removal setting, preview, and download. That is the same job as noise removal for remote meetings, webinars, and interview recordings—just applied to your lecture. Keep the original and compare before replacing the audio in your video.

ClearVoice is best for constant noise and quick cleanups. It will not fix a cough, a slammed door, or heavy room echo. For those, use the manual steps above.

Common mistakes that make lecture audio worse

Quick workflows by source

Zoom local recording

Find audio_only.m4a or the audio inside recording.m4a. If separate tracks were enabled, use your track. High-pass at 80 Hz, noise reduction 6–9 dB, normalize to -16 LUFS, export WAV.

Phone voice memo

Move the file to a computer. High-pass at 100 Hz for handling rumble. If you hear 50 or 60 Hz electrical hum, use a narrow notch filter at that frequency. For quick cleanup, use the same light noise-removal approach.

Webinar or interview with two speakers

Process each speaker separately. Learn a separate noise profile for each track. If you only have a mixed file, use less reduction—about 4–6 dB—so you do not pump the quieter speaker up and down.

Final checklist

  1. Keep the original untouched.
  2. Identify constant vs. intermittent noise.
  3. High-pass at 80–100 Hz.
  4. Sample 0.5–1 second of room tone.
  5. Apply 6–10 dB noise reduction, then check.
  6. Manually repair clicks, coughs, and gaps.
  7. Use de-reverb, gate, EQ, and compression sparingly.
  8. Normalize to -1 dB peak or -16 LUFS.
  9. Export WAV for video, 192 kbps MP3 for voice.
  10. Compare bypassed and processed versions on headphones and phone speaker.

You can rescue most lecture recordings. The trick is not one magic slider; it is a conservative chain: high-pass, manual edits, light noise profile, and a final listen. If the file is urgent, a free online pass can get you most of the way there. If you want maximum control, use Audacity or Audition and keep the settings gentle. Next time, record five seconds of room tone—it gives every noise-removal tool a better target.