In this guide
Find the low burst inside the complete word. Try a small local level reduction or a high-pass comparison on a copy, then listen to the whole sentence. Avoid deleting the consonant or shortening the timeline. A damaged word may need another take; a pop filter helps prevent the next recording problem.
To reduce plosives in a recorded voice, locate the low thump around a P or B sound and compare a small local level edit with a high-pass treatment. Keep the complete word and its original timing. The aim is a less distracting burst, not a missing consonant or a thinner voice throughout the recording.
A plosive can happen when a puff of air reaches the microphone. If that burst overloaded the recording or obscured the word, reducing its level cannot be assumed to recover the missing detail. Keep the original and check another take or microphone track before spending time on a badly damaged syllable.
Check whether the pop is actually a plosive
| What you hear | First check | Useful next step |
|---|---|---|
| A low thump tied to a spoken P or B | Listen to the complete word and the moment before it | Compare a local level dip and a low-cut treatment |
| A sharp tick at a new cut or join | Compare with the unedited source at the same time | Use the edit-click workflow |
| A gust or prolonged rush outdoors | Check the disturbance during speech, not only in a pause | Follow the wind-noise guide |
| Crackling throughout louder words | Check the original and an alternate recording | Treat it as possible recording damage rather than assuming every peak is a plosive |
| Pops only during one playback setup | Play the saved original through another available player or device | Investigate that playback path before processing the recording |
A large waveform peak is a location clue, not a diagnosis. Listen first, then zoom around that moment. The waveform reading guide explains why an overview cannot identify every sound or establish that a peak is clipped.
Lower one pop with a short, smooth level edit
- Make a working copy. Keep the original available on a muted reference track or as an untouched file. Include the whole sentence so you can judge the word in context.
- Locate the burst by ear. Loop a passage containing the complete word. Narrow the edit only after you know which moment is distracting; do not select the entire syllable just because it looks large.
- Draw a shallow dip around the burst. Use volume automation or an amplitude envelope with anchor points before and after it. Leave the surrounding voice at its previous level and make the transitions smooth.
- Compare from the same original. Start with a small reduction and increase only while the word remains natural. A deeper dip also lowers the wanted consonant and any vowel inside that interval.
- Listen outside the loop. Play the preceding phrase and the next few words. Undo an edit that creates an audible hole, a new tick or a missing word beginning.
Audacity's Envelope Tool reference documents control points for smooth volume changes. Its manual also warns that some effects can remove an envelope. If you combine filtering and an envelope, recheck the actual final export rather than assuming the dip survived the next processing step. Controls and behavior can vary by version.
Do not delete the burst's time interval from dialogue that must stay in sync with video. A level change can preserve timing; deleting time shifts what follows. Muting a whole word is also a different result from repairing its pop.
Compare a high-pass filter when the burst is mostly low-frequency
A high-pass filter progressively lowers frequencies below its cutoff. It affects the voice's low frequencies too, so it cannot identify which bass belongs to the pop. Audacity documents the effect under Effect → EQ and filters → High-pass filter, with cutoff and roll-off controls. Preview a conservative setting on a copy before applying it.
For a single pop, an editor that can process a time selection lets you compare a confined repair with filtering the whole sentence. Include the necessary consonant and check both selection boundaries: an abrupt change in tone can reveal the repair. There is no single cutoff or selection length that is right for every voice and burst.
The Saturalabs high-pass filter offers a separate browser comparison for supported short MP3/WAV recordings, without an account or credits. It applies a fixed 12 dB/octave slope to the entire opened recording; it has no time-selection repair or automation envelope. Use it to judge whether a low cut helps the overall recording, not to claim that only one P was repaired.
If a local repair is needed, finish it in an editor where you can retain the original timeline. Exporting a filtered excerpt and inserting it manually requires preserving its source position and duration. Review the audio replacement guide before returning edited dialogue to a video.
Why a gate, de-esser or AI voice target may leave the pop
A gate responds to level. A loud burst can open it, and a pop that coincides with speech can remain while the gate is open. A de-esser usually addresses harsh sibilant energy; it should not be assumed to solve a low air burst. Neither is a reason to remove the beginning of every word.
Voice isolation requests a source from a mix. A pop attached to that microphone's voice may remain with the requested speech. Saturalabs does not provide a dedicated plosive detector or guarantee reconstruction of damaged consonants. If you test its separate AI voice workflow for other competing sounds, compare the complete words in Isolated sound; AI WAV downloads require an account and credits. For one isolated pop, start with the local repair rather than assuming another separation pass is necessary.
Prevent plosives in the next take
A physical pop filter works before the air reaches the capsule. It cannot be added to an existing audio file. RØDE's microphone and pop-filter placement guidance recommends placing the filter between the speaker and microphone and experimenting with microphone distance and angle. Follow the orientation instructions for your particular microphone, especially whether it is addressed from the side or end.
- Record a short test containing the actual problem words, at the delivery level you will use.
- Compare the present position with a slight change in angle or distance, keeping other settings stable.
- Check quiet words, voice tone and room noise as well as the loud pop. A position that avoids the burst but makes every sentence distant may not be useful.
- Check input peaks and listen to the saved test before recording the full take. Lowering gain after distortion has already been captured does not restore the original waveform.
Check the exported repair before using it
Play the actual exported file with the original reference muted. Check that the complete word is intelligible, the surrounding voice has not become thin, and the repair has not added a click or a conspicuous level dip. For video, verify a visible speech cue after the repaired word and near the end of the clip. Preserve the original if a second attempt is needed.
If the word is still distorted or difficult to understand, an alternate take or a short re-recording may give a better result than stronger filtering. No before-and-after recording or measured repair result is claimed in this guide; the linked manufacturer and editor documentation explains the underlying controls and prevention methods.