How to apply a high-pass filter to voice recordings
- Step 1Confirm Pro access — Voice EQ (which contains the high-pass) is a Pro plan tool. Free locks it and caps the audio family at 50 MB / 30 min; Pro gives 200 MB / 120 min per file.
- Step 2Add the rumbly recording — One file per job. MP3, WAV, FLAC, M4A, OGG, and Opus decode. If you want to see the rumble first, run the spectrum-analyzer and look for energy piled up below 100 Hz.
- Step 3Pick the profile with the cutoff you want — Warm = 70 Hz (keeps the most low end), Male/Bright = 80 Hz (standard), Female = 100 Hz (most aggressive low cut, suits higher voices). Remember the profile also applies a low-mid cut and presence boost — there's no high-pass-only mode.
- Step 4Process in the browser — FFmpeg runs
highpassplus the profile's two EQ bands as one-afchain. Fast — seconds to under a minute — with a one-time WASM load on first run. - Step 5Verify the rumble is gone — A/B against the original on headphones, or re-run the spectrum-analyzer: the low-frequency shelf should be visibly rolled off. The voice itself should sound unchanged in body if you picked the right cutoff.
- Step 6Then normalise with confidence — With the rumble gone, loudness measurement is accurate. Finish on the loudness-normalizer for a clean -16 LUFS (podcast) or -14 LUFS (streaming) target.
High-pass cutoff by profile
The high-pass is the first stage of every profile. These are the exact cutoffs from the processor — there's no way to type a custom frequency.
| Profile | High-pass cutoff | Roll-off | Use when |
|---|---|---|---|
| Warm | 70 Hz | 2nd-order (12 dB/oct) | You want to keep maximum body; light rumble only |
| Male | 80 Hz | 2nd-order (12 dB/oct) | General-purpose broadcast standard for most voices |
| Bright (broadcast) | 80 Hz | 2nd-order (12 dB/oct) | Same cut as Male, plus the brightest tonal lift |
| Female | 100 Hz | 2nd-order (12 dB/oct) | Higher fundamental — the most aggressive low cut here |
What a high-pass removes from a voice track
The low-frequency problems this stage targets, and what survives.
| Noise source | Typical range | Removed by 70-100 Hz high-pass? |
|---|---|---|
| Desk thump / footsteps | 20-80 Hz | Mostly yes |
| HVAC / fan rumble | 30-120 Hz | Partly — the lowest partials |
| Mic handling / cable | 20-100 Hz | Mostly yes |
| 50/60 Hz mains hum (fundamental) | 50-60 Hz | Yes (the fundamental) |
| Mid-range hiss / fan whine | 1 kHz+ | No — use the ai-noise-reducer |
Tier limits
Pro feature; per-file; single file per job.
| Plan | Max size | Max duration | Access |
|---|---|---|---|
| Free | 50 MB | 30 min | Locked |
| Pro | 200 MB | 120 min | Yes |
| Pro-media | 100 GB | Unlimited | Yes |
| Developer | 100 GB | Unlimited | Yes |
Cookbook
High-passing rumble out of real recordings, with the exact FFmpeg chain each profile runs.
Desk-thump in a home-office interview
Male profile's 80 Hz high-pass is the standard low cut and removes desk and chair thumps. (The profile also cuts 300 Hz and boosts 3 kHz — there's no high-pass-only mode.)
Profile: Male highpass=f=80, <- the low cut equalizer=f=300:t=q:w=1:g=-2, equalizer=f=3000:t=q:w=1:g=3 interview.wav -> interview-eq.wav
Higher voice with traffic rumble through a window
Female profile applies the most aggressive cut at 100 Hz, suited to a higher fundamental, taking out more of the low traffic energy.
Profile: Female highpass=f=100, <- most aggressive low cut equalizer=f=300:t=q:w=1:g=-2, equalizer=f=4000:t=q:w=1:g=3
Deep voice — keep body, cut only the rumble
Warm's 70 Hz high-pass is the gentlest, preserving a deep voice's natural low end while still trimming sub-voice rumble.
Profile: Warm highpass=f=70, <- keeps the most body equalizer=f=350:t=q:w=1:g=-1, equalizer=f=2500:t=q:w=1:g=2
See the rumble before and after
Confirm the low-cut worked by viewing the spectrum.
1. /audio-tools/spectrum-analyzer on raw file -> big shelf of energy below 100 Hz 2. voice-eq (Male) -> high-pass at 80 Hz 3. /audio-tools/spectrum-analyzer on output -> low shelf rolled off
High-pass won't touch mid hiss
A high-pass only addresses low frequencies. Fan whine, hiss, and hum partials up in the mids need a different tool.
Low rumble -> voice-eq high-pass (this tool)
Hiss / mid-range fan noise / steady hum
-> /audio-tools/ai-noise-reducer (RNNoise)Edge cases and what actually happens
Wanted high-pass only, no other changes
Not availableThere's no high-pass-only mode. Every profile bundles the high-pass with a low-mid cut and a presence boost in one chain. Pick the profile whose extra bands you can live with — Warm changes the tone the least.
Can't set a custom cutoff frequency
By designThe cutoff is fixed per profile: 70 Hz (Warm), 80 Hz (Male/Bright), or 100 Hz (Female). There's no field to type your own. For an arbitrary cutoff, you'd need a DAW or a manual FFmpeg command.
Rumble still audible after high-pass
PartialA 2nd-order (12 dB/oct) roll-off attenuates but doesn't brick-wall. Strong rumble whose energy reaches up to 120-150 Hz won't be fully removed by an 80 Hz cut. Try Female (100 Hz) for more aggressive cutting, or treat the source.
Voice sounds thinner after high-passing
Cutoff too highIf a deep voice lost body, the 100 Hz (Female) cut was too aggressive for it. Use Warm (70 Hz) or Male (80 Hz) to keep more low end while still trimming rumble.
Hiss or fan whine remains
Wrong toolA high-pass only addresses low frequencies. Broadband hiss and mid-range fan noise need the ai-noise-reducer, which uses an RNNoise speech model rather than a frequency cut.
Mains hum still present
PartialThe fundamental of 50/60 Hz hum is removed by the high-pass, but its harmonics (100, 120, 150 Hz…) sit above the cutoff and survive. For stubborn hum, denoise as well.
Free user can't run it
Pro requiredGated at Pro despite 'free' in the URL. Free locks the tool and caps the audio family at 50 MB / 30 min.
File over the duration limit
400 — duration exceededPro allows 120 minutes per file. Longer field recordings are rejected even under 200 MB — split with the audio-splitter, process, then rejoin.
MP3 in, slight re-encode loss
ExpectedEQ re-encodes to the same container. MP3 in becomes a re-encoded MP3 out. Process a WAV/FLAC master if you have one.
Sub-bass content you wanted to keep
RemovedIf the recording legitimately has wanted low-frequency content (e.g. a music bed with deep bass), the high-pass will roll it off too. EQ the voice track separately from the bed.
Frequently asked questions
How do I high-pass filter a voice recording?
Pick a Voice profile — each starts with a high-pass: 70 Hz (Warm), 80 Hz (Male/Bright), or 100 Hz (Female). The filter rolls off rumble below the speaking range as the first stage of the chain.
Can I apply only a high-pass with nothing else?
No. The high-pass is always bundled with that profile's low-mid cut and presence boost. There's no high-pass-only mode here — Warm changes the tone the least if you want the smallest extra effect.
Can I choose the cutoff frequency?
Only by profile: 70, 80, or 100 Hz. There's no custom-frequency field.
How steep is the filter?
It's FFmpeg's highpass at the default 2nd-order, roughly 12 dB per octave — enough to kill rumble without thinning the voice.
Will it remove AC hum?
It removes the low fundamental (50/60 Hz) but not the harmonics above the cutoff. For full hum removal, also run the ai-noise-reducer.
Will it remove hiss?
No — hiss is mid-to-high frequency. A high-pass only touches the low end. Use the ai-noise-reducer for hiss.
Is my recording uploaded?
No. It runs in your browser via FFmpeg WASM and stays on your device.
Is it free?
Free of upload and cloud cost, but the tool requires a Pro plan. The Free tier doesn't include it.
Which cutoff should I use?
80 Hz (Male) is the safe default. Use 70 Hz (Warm) to keep more body on a deep voice, and 100 Hz (Female) for the most aggressive cut on a higher voice.
What formats can I use?
MP3, WAV, FLAC, M4A, OGG, and Opus. The output keeps the input format.
Can I high-pass several files at once?
No — one file per job. For a multi-file chain, use podcast-master.
How can I see if the rumble is gone?
Run the spectrum-analyzer before and after — the energy below 100 Hz should be visibly reduced after the high-pass.
Privacy first
Every JAD Audio tool runs entirely in your browser via FFmpeg (WebAssembly) and RNNoise. Your audio files never leave your device — verified by zero outbound network requests during processing.