Low-frequency rumble
Compare a low cut when a deep background sound is distracting. Listen for a thinner voice, lost bass notes or an altered attack. This filter cannot distinguish unwanted rumble from wanted sound at the same frequencies.
Free tools for your next audio edit
Free high-pass filter. Test a low cut before stronger cleanup.
Reduce low-frequency rumble in a short MP3 or WAV. Set a cutoff, apply a fixed 12 dB/octave filter locally, then compare the actual export with your original. No account, credits or audio upload. Low-cut filtering can change voices and instruments too.
No account neededNo audio uploadWorks in your browser
Choose a repair for wind and rumble High-pass filter audio
Free · localNo file selected. Audio stays in this browser.
Applies one 12 dB/octave high-pass filter independently to each channel. No normalization, trimming, source separation or reconstruction of damaged words. Filter phase and startup transients change the waveform; peaks can rise. Download before leaving and keep your original.
Compare the voice as well as the rumble
How to high-pass filter an MP3 or WAV
Keep the original. A lower noise level is useful only if the sounds you need still work in the edit.
- 01 / OPEN
Choose a representative recording
Open a mono or stereo MP3/WAV up to 50 MiB and five minutes. Include the low sound, a quiet word and an important attack. The browser decodes it locally at 48 kHz.
- 02 / SET
Choose a cutoff to compare
Start from the displayed 80 Hz or another whole-number setting from 20 to 1,000 Hz. The slope is fixed at 12 dB/octave. Higher settings remove more low-frequency material, including wanted voice body or musical bass.
- 03 / REVIEW
Prepare and listen to the actual export
Choose WAV or MP3, apply the filter and compare the two players. Download the same file used by the result preview. Each new setting starts from the selected original; it does not stack filters on the previous result.
A frequency-based low cut
A high-pass filter reduces low frequencies across the mix.
High-pass and low-cut describe the same operation here: lower frequencies are progressively attenuated while higher frequencies pass more readily. A cutoff is not a wall. Sound below it is reduced by an amount that depends on frequency; it is not automatically deleted.
This tool uses one second-order Butterworth high-pass biquad per channel, with Q = 1/√2. Coefficients follow the high-pass equations in the W3C Audio EQ Cookbook. The filter is causal, starts with zero internal state and changes phase. It is not zero-phase filtering, a noise profile, a gate or a source-separation model.
The same coefficients are used on both stereo channels, with independent processing state. A sound on the left is not mixed into the right. The complete decoded frame count remains, without trimming, time stretching, automatic edge fades or loudness normalization.
Choose by the problem you hear
Low rumble, quiet gaps and overlapping sources need different tests.
Wind during a sentence
Some wind extends beyond the lowest frequencies or damages the captured voice. Use the wind-noise repair guide to choose between filtering, voice isolation and another take. A low cut cannot guarantee recovery of missing words.
Noise between phrases
The free noise gate lowers quiet passages by level while keeping their duration. That is different from cutting low frequencies throughout a recording. A gate can still leave noise beneath an open, louder voice.
Keep an honest comparison
Filtering is not bass-instrument separation.
A bass guitar, kick drum and low voice can share frequencies. A high-pass filter changes all of them according to frequency. If you need an estimated musical source rather than a frequency range, the bass extraction workflow is a different task with its own account and credit requirements.
Compare at a similar listening level and review the whole passage before increasing the cutoff. If the voice loses useful body before the rumble is acceptable, retain the original and try another method. The default is a starting point, not a universal speech preset.
Peaks can rise after filtering even though low frequencies are reduced. Out-of-range prepared samples are clipped during PCM16 conversion and counted in the result. No limiter is applied. If the count is nonzero, reduce the original level in an editor before testing again; turning down the already clipped download cannot restore its peaks.
Local processing, real file preview
Know what the WAV or MP3 download contains.
WAV exports 48 kHz, 16-bit PCM with the decoded mono/stereo layout and frame count. It must fit 50 MiB. MP3 applies one new lossy encode at 128, 192 or 320 kbps, with the actual file subject to the same byte ceiling. MP3 can add encoding delay or padding and can change peaks after decoding; the clipping count refers to PCM preparation, not the final decoded MP3.
Original compressed bytes, tags, artwork and bit depth are not copied. Download before leaving or reloading the page. For a video, retain the original source start, use WAV for the working replacement, and check the exported picture against the audio replacement and sync guide.
No audio is uploaded and no AI processing job starts. The loudness meter can inspect the chosen working WAV after processing; this filter does not certify a delivery loudness or true-peak limit.
One known input, one real filter output
Hear a low cut on two synthesized tones.
A low 25 Hz tone accompanies a higher 440 Hz tone. Compare the original with this tool's actual 80 Hz export at the same player level. These files demonstrate frequency filtering, not wind recognition, speech recovery or AI separation. Some speakers do not reproduce the lowest tone usefully.
Original: low tone and higher tone
A 25 Hz tone at amplitude 0.18 accompanies a 440 Hz tone at amplitude 0.12, with 20 ms outer fades. This is synthetic audio, not wind or speech.
Download the originalFiltered: 80 Hz high pass
The actual local filter reduces the low tone while mostly retaining the higher tone's steady-state amplitude. Phase and outer-edge behavior change; no normalization is applied.
Download the filtered WAV| Tone | Original amplitude | Filtered amplitude | Change |
|---|---|---|---|
| 25 Hz | 0.179997 | 0.017494 | -20.247 dB |
| 440 Hz | 0.119999 | 0.119931 | -0.005 dB |
Both files have 192,000 mono frames, 4 seconds, 48 kHz, 16-bit PCM and 384,044 bytes. The filter reduces rather than deletes the low tone, changes phase and does not normalize the result. These are tone-amplitude readings, not loudness, SNR or percentages of wind removed.
To repeat the test above, download the original, choose it in the editor, leave 80 Hz and WAV selected, and apply the filter. Changing the cutoff starts another comparison from that original. For more reference signals, use the WAV test-file library.
These original learning assets are dedicated under CC0 1.0. Read the generation and verification notes or download the measurement report. The stored sample comparison is not a test of your player or device.
MP3 encoder credits
Built with the LAME encoder via lamejs.
The optional MP3 download uses lamejs 1.2.1 by Alex Zhukov, a JavaScript port of the LAME MP3 encoder. Its unmodified script is served separately. See the attribution and source notes, upstream license notice, LGPL-3.0 and GPL-3.0 texts, and complete upstream source package.
A few useful answers
Free high-pass filter questions
How do I apply a high-pass filter online?
Choose a supported mono or stereo MP3/WAV up to 50 MiB and five minutes. Set a whole-number cutoff from 20 to 1,000 Hz and choose WAV or MP3. Apply the filter, compare its actual preview with the original, and download that same file. Processing and downloads are free and local.
What does a high-pass filter remove?
It progressively reduces frequencies below the cutoff. It does not identify wind, voices or instruments. Wanted bass, voice body and other low-frequency material are affected too. It is not a brick-wall filter and does not remove every sound below the chosen number completely.
Is this a low-cut filter or bass remover?
High-pass and low-cut describe the same direction of filtering here. It can reduce low-frequency energy across the whole mix; it does not separate a bass instrument from the other sounds. For a musical source, use a source-isolation workflow rather than assuming EQ recognizes the instrument.
What cutoff should I use for wind rumble?
The initial 80 Hz is a starting point to compare, not a universal speech or wind preset. Test the gust and quiet words, and lower the cutoff if the voice loses too much body. Wind can extend beyond the lowest frequencies, and damaged words may remain damaged. Keep the original for every new test.
Does filtering preserve timing and stereo?
The filter retains the decoded frame count and processes left and right independently with the same coefficients. It is a causal filter with phase change and zero initial state, not zero-phase processing. Startup transients and changed peaks are possible. MP3 adds a lossy encode and can add delay or padding; check a video replacement in the editor.
What slope and download formats are available?
The slope is fixed at 12 dB/octave, using a second-order Butterworth high-pass filter at 48 kHz. WAV writes 16-bit PCM; MP3 uses 128, 192 or 320 kbps. The WAV and actual MP3 must fit 50 MiB. Original tags, bit depth and compressed bytes are not preserved.
Will the filter normalize levels or fix clipping?
No. It applies no normalization, limiter, edge fade or clipping repair. Peaks can rise after filtering. PCM preparation clips out-of-range samples and reports their count; reduce the original source level and retry if needed. That count is not a true-peak or decoded MP3 peak measurement.
After preparing your audio