How to Remove Mouth Clicks Breaths and Lip Smacks from Voiceovers Without Losing Clarity

Clean voiceovers start with capture and finish with selective editing: keep the mic about 6 to 12 inches from your mouth, slightly off-axis, aim for peaks around -12 dB to -6 dB, then remove clicks first, reduce breaths second, and use broad noise processing last.

If your voiceover sounds fine on speakers but suddenly feels full of tiny clicks and sharp breaths in headphones, you are not imagining it. Small mouth noises become much more obvious in short-form videos, product demos, lessons, and captioned social clips where the voice is dry, close-miked, and heavily exposed. This guide gives you a practical cleanup workflow, starting with prevention, then moving into fast edit decisions that keep the track natural.

Know Which Noise You Are Hearing

Mouth clicks are short, sharp transients caused by saliva, tongue release, or teeth and palate contact. Lip smacks are slightly longer adhesive sounds made when the lips separate. Breaths are inhalation or exhalation noises between phrases, and unlike clicks, they are not always a problem. The first decision is not "remove everything." It is "remove what distracts from the line."

For creator workflows, the real target is intelligibility and listening comfort, not a clinically sterile waveform. A platform describes high-fidelity audio as clean, accurate sound with minimal noise or distortion while preserving the natural dynamics of speech. Clearer audio also improves comprehension, reduces listener fatigue, and supports accessibility, which matters for education videos, product explainers, and social clips where viewers may already be splitting attention between voice, captions, and visuals.

Prevent More Than You Fix

Most cleanup time is created before you open the editor. A quiet room, soft furnishings, a pop filter, over-ear monitoring, and mic placement slightly off-axis do more for mouth-noise control than heavy processing after the fact. Internal phone microphones can work if the phone stays close to your mouth and the room is quiet, but quality drops as distance increases. USB microphones are easy to deploy, but they often give you less control and can introduce more noise or distortion than a dynamic XLR microphone paired with an interface.

Dryness matters too. A study indexed on a platform examined whether hyposalivation, or oral dryness, affects women's voice quality. In practice, that means mouth clicks are not just a mic problem; they are often a performance-state problem. For voiceover sessions, use a short warm-up, avoid starting cold, keep water nearby, pause between takes instead of powering through a dry read, and record three short test lines before the real take. If the first sentence already sounds tacky or sticky in headphones, fix the setup before recording the full script.

Use a Cleanup Order That Preserves Speech

A reliable voiceover cleanup order is: 1) duplicate the raw track, 2) remove mouth clicks and lip smacks, 3) rebalance breaths, 4) clean room noise lightly, 5) apply EQ and compression, and 6) generate captions from the cleaned file. This order matters because clicks are surgical problems, while noise reduction, EQ, and compression are broad processes. If you compress first, the clicks and breaths often become louder relative to the words.

Use this checklist on every pass:

1
1. Solo the voice track and listen once at normal speed.
2
2. Mark obvious clicks, lip smacks, and intrusive breaths before editing anything.
3
3. Fix the shortest transients first with cuts, clip gain, or spectral repair.
4
4. Reduce breaths by level before deleting them entirely.
5
5. Apply noise reduction only after the edit pass, and stop if consonants start to sound dull or watery.
6
6. Generate captions and timing only after the spoken track is final.

In CapCut or a similar AI-assisted editor, this usually means doing the surgical edits first, then using AI tools for repetitive tasks such as rough leveling, background cleanup, or captions. AI can speed up the workflow, but it should support review, not replace it. If a tool removes detail from consonants, shifts a word boundary, or makes the voice sound phasey, roll it back and fix that spot manually.

Remove Each Noise Type the Right Way

A good rule is to match the tool to the duration of the problem. Very short noises respond best to tight edits and fades. Longer or more complex noises may need clip gain, replacement room tone, or spectral repair. Use these starting ranges as workflow defaults, not hard rules.

Mouth Clicks

Mouth clicks are usually the easiest to over-process. Start by zooming into the waveform and listening to the click in isolation. If it sits between words, split just around the transient, reduce the clip gain by 8 dB to 20 dB, and add very short fades. If it overlaps the start of a consonant, use a de-click or spectral-repair tool on that tiny region only. A single low- to medium-sensitivity pass is usually safer than global de-clicking across the whole take.

Breaths

Breaths carry pacing and can make a read feel human, so default to reduction instead of deletion. For tutorials, explainers, and education videos, start with -3 dB to -6 dB. For ads, reels, and tight product promos, -6 dB to -12 dB is a better starting range. Remove a breath entirely only when it lands against a hard cut, a music hit, or a subtitle change that makes the inhale feel louder than the words.

Lip Smacks

Lip smacks are longer and stickier than clicks, so they often need a slightly wider edit. If the smack is between phrases, cut it and bridge the gap with short crossfades. If it sits under the start of a word, reduce the smack first and preserve the spoken onset. When the artifact is broad or textured, replace the section with matched room tone rather than forcing heavy de-clicking. If the repair makes the voice sound processed, the manual edit is usually the better choice.

Adjust the Cleanup by Content Type

Different video formats need different breath policies because the voice carries different jobs. Social ads and short reels are rhythm-driven, so stray mouth sounds feel larger against quick cuts, motion graphics, and dense captions. Education content is more tolerant because the listener expects a continuous explanation, not a hyper-compressed promo rhythm. Product demos sit in the middle: clean enough to feel polished, but not so stripped that the narrator sounds synthetic.

Use case should drive the breath target:

Clearer audio improves comprehension and reduces listener fatigue, so the stricter your format is about pace and attention, the less tolerance you have for random mouth noise. For social packaging, do the voice cleanup before captions and timing polish. For longer-form lessons, keep a few quiet breaths if they support phrasing and prevent the read from sounding machine-cut.

Final Takeaway

The best way to remove mouth clicks, breaths, and lip smacks is not one big AI cleanup button. It is a sequence: record cleanly, fix the shortest noises first, reduce breaths instead of deleting them by default, and apply broad processing only after the surgical pass. If a repair makes speech less clear, the fix is worse than the noise.

Use this repeatable session workflow:

1
1. Record in a quiet, soft room with the mic slightly off-axis.
2
2. Keep levels conservative and leave headroom.
3
3. Mark clicks, smacks, and heavy breaths on the first listen.
4
4. Edit clicks and smacks manually or region by region.
5
5. Reduce breaths by use case, not by habit.
6
6. Apply light EQ, compression, and noise reduction last.
7
7. Review once on headphones and once on phone speakers before exporting.

FAQ

Q: Should I Remove Every Breath From a Voiceover?

A: No. In most creator workflows, reducing breaths sounds better than deleting every one. Full removal often makes narration feel jumpy or unnatural, especially in lessons, explainers, and testimonial-style reads.

Q: What Is the Fastest Way to Remove Mouth Clicks?

A: Start with isolated manual fixes. For short clicks between words, split the transient, lower the clip gain, and add tiny fades. Use de-click or spectral repair only on the affected region first, because full-track de-clicking can soften consonants.

Q: Why Do I Hear More Mouth Noise on Short-Form Videos Than on Long Podcasts?

A: Short-form edits expose the voice more aggressively. Closer miking, louder mastering, tighter gaps, music hits, hard cuts, and large captions all make tiny transients feel bigger, so click and breath control has to be more deliberate.