How to Remove Hum from Audio Online

What hum actually is

Creators call every unwanted sound static, but hum has a recognizable pattern: a low, steady mmmm tone, usually electrical, often from a ground loop, a power adapter, USB mic noise, or interference. It is concentrated around 50 Hz or 60 Hz and may carry harmonics above that base.

iZotope describes hum removal as targeting a base frequency, usually 50 or 60 Hz, plus harmonics. Audacity's manual notes noise reduction can reduce constant hum but recommends notch filtering before noise reduction for mains hum.

Hum vs buzz vs hiss vs rumble

Naming the problem helps you pick the right fix.

• Hum — low steady tone from ground loop or power. AI cleanup first; notch if tonal.
• Buzz — sharper raspy electrical sound from bad cables or dimmers. AI cleanup; spectral denoise or de-hum if tonal.
• Hiss — high ssss from preamp gain or cheap mic. AI denoise or noise-profile reduction.
• Rumble — low vibration from HVAC, fridge, or traffic. High-pass filter plus AI cleanup.

The AI-first workflow

For most creator recordings, AI cleanup is the fastest first pass. Upload the file, run a light pass, then preview around a quiet pause, a soft-consonant sentence, and a louder sentence where the hum sat under the voice.

Keep the original. Clean before compression, because compression raises quiet noise and can make hum harder to remove. For video, clean the audio before captions and translation so every downstream step starts from clearer speech.

Hum diagnosis checklist

Answer these before overprocessing: Is the noise constant? Is it a single low note, suggesting 50/60 Hz electrical hum? Is it a raspy tone with higher overtones, suggesting buzz from lighting or motors? Does it disappear when you unplug the charger? Does it get louder when the AC or fan runs? Is the voice barely louder than the hum?

1. The noise is constant across the take — a cleanup pass is worth trying.
2. It is a single low note near 50 or 60 Hz — a notch may finish it after cleanup.
3. It is a raspy tone with overtones — suspect lighting, motors, or cables, not mains hum.
4. It changes when a cable or charger moves — fix the source before processing.

AI vs notch vs high-pass vs gate

Online AI cleanup suits steady hum, fan, and HVAC noise for non-technical creators. A notch filter suits a pure 50/60 Hz electrical tone and its harmonics. A high-pass filter suits low rumble from HVAC and handling. A noise gate only silences between phrases; Audacity defines it as allowing audio above a threshold to pass, so it cannot fix hum under speech.

Fixing stubborn electrical hum

Use the notch only if AI cleanup leaves a clear tonal whine. Find the likely base frequency first: 60 Hz in the US, 50 Hz in many other regions, and possibly off the exact number for resampled or speed-changed files. Look for harmonics at multiples like 120 Hz, 180 Hz, and 240 Hz, shown as horizontal lines in a spectrogram.

Apply narrow cuts at the base and one or two obvious harmonics. Do not notch every multiple: too many cuts thin the voice, especially in lower voices. Then run a lighter AI pass and compare against the original.

Hum prevention

Prevention beats repair. Turn off or move away from refrigerators, AC units, fans, and computer towers. Put the mic closer to the speaker and farther from machines. Test a different cable, port, charger, or outlet. Use balanced XLR or TRS connections when your gear supports them, and keep audio cables away from power bricks.

Processing order for a mixed-noise recording

Many recordings carry more than one problem: a low hum from the desk plus keyboard clicks, or an AC rumble plus room echo. The safe order is to fix the broad bed first, then the tone, then the events. An AI cleanup pass reduces the steady layer; a notch handles the tonal hum that survives; manual edits catch a click or a cough. Each step should be judged separately, because a change that helps the hum can make a click more noticeable.

Cleanvoice's own cleanup guides follow the same structure: name the dominant sound, apply the lightest matching process, and review during speech rather than only in silence.

How long a typical cleanup takes

A browser pass on a normal-length file finishes in a minute or two of processing. The listening check adds a minute more. A full DAW session with a notch and a gate can stretch to twenty minutes once you are checking several passages. The time difference is why a quick pass first is usually the right first move, with manual fixes only for what survives.

What to check before you pay

If a tool offers a free preview, use it before any purchase. Listen to the same three passages in the preview: the quietest pause, a consonant sentence, and the loudest section where the hum sat. If the preview preserves the voice, the paid version will too. If the preview already sounds thin, no setting change will add back what the source never had.

How loud should the notch be?

A notch is strongest when it is narrow. Cut only the base frequency and at most one or two clear harmonics, by a modest amount. A wide, deep cut at 60 Hz often removes low-mid body from a voice that already sits in that range. The hum is reduced; the voice gets thinner. If after a narrow cut the hum is still audible, address the source (cable, charger, grounding) rather than repeating the cut deeper.

A voice note on 50 Hz vs 60 Hz

Which frequency you check depends on where you recorded. The U.S. and parts of Asia run 60 Hz; much of Europe, Latin America, and Africa run 50 Hz. A file recorded from an old tape, a speed-changed video, or a resampled export may sit slightly off the exact value, which is why a spectrogram helps: the hum appears as a horizontal line you can read directly instead of guessing the number.

Understanding the AI-first order

AI cleanup first works because the model reduces the whole steady bed, including rumble and buzz that are not a single tone. Once the bed is lower, the remaining tonal hum is easier to hear and to notch without over-cutting. The reverse order — notching first — risks cutting a tone you did not need to cut, then applying a stronger cleanup than the file needed.

What counts as a successful hum cleanup

Success is a recording where the low tone stops competing with the voice and the speech keeps its natural weight. A hum that was masking a softer consonant becomes less obtrusive; the pause stops vibrating. The opposite of success is a voice with the low-mid removed and the hum still present. The difference between the two is usually a lighter setting and one narrow cut, not a stronger one.

Hum in a final mix

Hum reduction also changes the loudness stage. Once the low bed is gone, normalization and compression can be set from the voice rather than from the hum, and the episode can settle at the same level every week. For a series, a repeatable cleanup followed by a repeatable loudness check is what makes episode forty sound like episode four. The cleanup is the first step in a chain, and the chain is what creates consistency.

SimpleClean vs Audacity vs Adobe Audition

SimpleClean suits creators who want hum removed quickly without a DAW. Audacity suits free manual editing with noise profiles and notch filters. Adobe Audition suits professional restoration with spectral display and DeHummer-style workflows. For a severe electrical fault or hum louder than speech, fix the source and re-record.