Skip to content
Neattrim

Remove fan hum from a recording

· Neattrim editorial team

Do it now
Noise remover
Remove background noise from speech.

Steady fan noise is a better candidate for noise reduction than an unexpected door slam over a word. A fan may produce both a low tonal hum and broad rushing air. Those components can need different treatment, and pushing one process harder may damage speech before removing both.

Identify the noise you actually have

Listen during a pause and during a sentence. A stable low tone suggests hum; a wide hiss or whoosh suggests airflow. Rumble from handling or desk vibration is another problem. Room echo follows speech and is not the same as a steady noise floor.

If you can record again, moving the microphone closer to the speaker and away from airflow often helps more than repairing the export. Do not cover a device's ventilation to quiet it. Try a different microphone position or a quieter recording location.

Make a conservative first pass

  1. Keep the original and select a passage containing speech plus a pause.
  2. Process a copy with Noise Remover.
  3. Compare the quiet pause, soft consonants and sentence endings with the original.
  4. Keep the result only if the words remain natural and easier to follow.

A quieter pause alone is not enough. If the speaker becomes watery or words disappear, the reduction is too destructive. Accepting a little residual noise can produce a more usable recording than forcing silence.

Filters and models have different strengths

A notch filter can reduce a narrow tone, but also removes useful audio in that region. A high-pass filter can reduce low rumble while thinning a voice if pushed too far. Spectral denoisers estimate noise across frequency bands; learned speech-enhancement models use patterns learned from training data.

The FFmpeg documentation for audio filters, checked on 2026-11-13, describes spectral denoising and filtering approaches. These are processing options, not a guarantee that the web tool uses a particular model or exposes those settings.

Avoid making the next stage worse

Boosting volume after denoising also raises whatever noise remains. Repeated cleanup passes can accumulate artifacts, especially when each pass treats remnants of the previous pass as new noise. Return to the original when trying a different approach.

Listen on headphones and the intended playback speaker. Check breaths and quiet endings, not just the loudest sentence. For interviews, make sure a quieter second speaker has not been suppressed as background.

When to stop

If noise masks a word completely, no process can guarantee the correct original word. Re-record important material when possible, or request clarification rather than treating a plausible-sounding reconstruction as evidence. Once speech is clean enough, adjust its level carefully.

Related guides