If you typed how do I get rid of background noise into a search bar, you are probably staring at a good take and an ugly room. It might be a podcast, Zoom call, phone memo, gaming clip, or client video. You can often rescue it without re-recording, but no single slider fixes everything. Use this order.

Step 1: Identify the noise

Listen on headphones. Different noises need different tools.

Step 2: Make a safety copy

Duplicate the original. If you have video, export the audio as WAV 48 kHz 24-bit or keep the original MP4 until final. Do not bounce between MP3 and AAC repeatedly; compression artifacts get amplified by noise removal.

Step 3: Run AI noise removal first

For steady noise, AI is the fastest first pass. It separates voice from fan hum, road noise, hiss, and room tone better than a gate. ClearVoice is a free online option for AI background noise removal for video and audio. Upload the file, choose voice cleanup, set strength to medium or about 60 to 80 percent, and process. The tool accepts MP4, MOV, WAV, MP3, and M4A, so you can clean a video clip without separating audio first.

Listen. If the voice sounds underwater or metallic, lower the strength. If hum remains, keep the AI result and fix it with EQ. AI handles constant noise well, clicks decently, and sudden barks or overlapping speech poorly.

Step 4: Clean leftovers with EQ

  1. High-pass filter: 80 Hz for deep male voice, 100 Hz for higher male or female, 120 Hz if rumbly but thin. Removes traffic, footsteps, desk thumps.
  2. Notch for hum: narrow notch at 50 or 60 Hz, Q 10 to 20, cut 12 to 20 dB. Use 60 Hz in the US, 50 Hz in many other countries.
  3. Boxiness: cut 2 to 4 dB around 250 to 400 Hz. More than that makes the voice thin.

Solo the track as you cut so you hear only the voice.

Step 5: Gate or expand the pauses

A gate mutes when you stop talking, but hard gates chop word tails and pump. A downward expander is gentler. Start at 2:1 or 3:1, threshold just above the noise floor, attack 5 to 10 ms, hold 50 to 100 ms, release 100 to 200 ms. If using a hard gate, try -40 to -30 dB. It should close in silence and open instantly on speech. Watch the waveform: the gate should close only in the quiet gaps, not mid-word.

Step 6: Fix intermittent noises by hand

If the background is music or TV, AI may leave artifacts. Re-record or cut around it instead of forcing it.

Step 7: Treat echo separately

AI noise removal is not de-reverb. If you sound like you are in a bathroom, use a de-reverb tool or record again. Next time, get closer to the mic: 3 to 6 inches for a dynamic, 6 to 12 for a condenser. A blanket behind you and a rug under you help more than plugins.

Step 8: Export without undoing the cleanup

Audio: WAV 48 kHz 24-bit, or MP3/AAC at 320 kbps. Video: H.264 MP4 with AAC 320 kbps at 48 kHz. Aim for -16 LUFS integrated for mono podcast and -14 LUFS for stereo YouTube or Spotify. Keep true peak at -1.5 dBTP or lower. Do not normalize to 0 dB. If the file is mono, keep it mono; fake stereo wideners can make noise worse.

Step 9: A/B test on real devices

Bypass the processing and compare. Phone speakers expose pumping and harshness. Earbuds expose metallic artifacts. If word tails chop, reduce the gate. If the voice warbles, reduce AI strength. If room tone disappears and reappears, use an expander instead of a gate.

Step 10: Prevent it next time

The 10-minute order of operations

  1. Duplicate the file and export a clean WAV or keep the original video.
  2. Run AI noise removal first. Try ClearVoice for a free online option.
  3. High-pass filter at 80 to 120 Hz. Notch 50 or 60 Hz if hum remains.
  4. Use an expander or gentle gate to quiet pauses.
  5. Patch clicks, barks, and sirens manually.
  6. Export at 48 kHz with true peak below -1.5 dBTP.

That is the practical answer to how do I get rid of background noise. Remove constant noise with AI, clean the low end with EQ, control pauses with an expander, and patch weird noises by hand. Done in that order, a noisy voice memo or Zoom call can sound like it came from a much better room.