RX 12 Mouth De-click: Remove Lip Smacks Safely

RX 12 Mouth De-click reduces distracting saliva clicks, tongue noises and lip smacks in speech and vocals. Start with a representative sentence, raise Sensitivity only until the obvious clicks are caught, then use the Listen ear button during Preview in the RX 12 Editor. If useful consonants, attacks or syllable edges enter the removed signal, back the setting down. Return to normal speech and approve it before committing the repair.

Frequency Skew shifts detection toward lower or higher click energy. Click Widening extends the repair around a detected event, which helps a lip smack with a short tail but can also erase more wanted sound. Use a conservative whole-file pass for dense click fields and spot-repair the stubborn events rather than forcing one aggressive preset to solve everything.

What Mouth De-click is designed to remove

The current RX feature comparison lists Mouth De-click for clicks and lip smacks. A click can appear as a short vertical spectrogram trace; a wet smack may have a longer tail. Those shapes are clues, not a diagnosis: wanted consonants can look similar. Close capture and later bright EQ or compression can make small mouth sounds easier to hear.

Mouth De-click is available as both a module and a plug-in in RX Standard and Advanced. That gives you two valid routes: process a selected range in the Audio Editor or run a conservative real-time instance in a DAW/NLE. Elements includes the general De-click plug-in, not the dedicated Mouth De-click component.

For speech-specific lip and saliva noises, start here. The RX De-click tutorial covers digital discontinuities, vinyl clicks and other general impulses, while the vocal cleanup chain places mouth repair beside plosives, breaths, noise and reverb.

Choose the Editor or plug-in route

The nine steps below use the standalone RX 12 Audio Editor. Its Preview, Compare and Render controls are different from a real-time insert: in a DAW, start playback in the host and use its bypass and bounce controls. AudioSuite uses its own host preview/render controls. The Editor’s Listen ear button auditions removed audio during Preview. The footer is absent in Module Chain; audition the individual module outside the chain. RX plug-ins also omit the Editor footer controls, except for Bypass. If your plug-in offers Output Clicks Only, use that monitor during host playback and turn it off before bouncing. Check the actual plug-in window rather than looking for the Editor footer.

Real-time processing is not a promise of zero monitoring delay. Check the host’s delay compensation and the rendered file’s start, end and alignment. The RX plug-in latency guidance explains how delay is reported to a compatible host. For live recording, keep an untreated take and test the monitoring path before committing the repair.

In the Editor, work on the intended file tab with Composite View off and Instant Process off while making selections. Standard handles mono/stereo files; use the Advanced Editor for supported multichannel files rather than folding channels just to run this repair. Confirm the plug-in format and channel configuration in your actual host separately.

Mouth De-click versus De-click

They overlap. The RX 12 De-click manual explicitly lists its Low Latency algorithm as useful for mouth clicks. Mouth De-click offers a dedicated detector with three main adjustments; general De-click offers selectable algorithms. Neither is guaranteed to win on every voice. Do not stack them automatically; audition the remaining defect before adding another pass.

A low thump on a P or B is a plosive, not merely a mouth click. Treat the low-frequency burst with De-plosive, then address any separate high-frequency lip noise. iZotope's audio-artifact guide recommends that division for difficult plosive-plus-mouth-noise events.

Read the three Mouth De-click controls

Sensitivity determines how many candidate mouth clicks are detected. Higher is not “better.” The RX 12 Mouth De-click manual warns that excessive sensitivity can affect plosives and damage wanted signal. Set it by what disappears from the voice, not by the repaired-click count.

Frequency Skew biases detection toward lower or higher frequencies. The documented guidance says negative values suit more generic low clicks, while zero and positive values target mouth clicks in the middle frequencies. Use the spectrogram and removed-signal monitor instead of assuming every speaker needs a bright setting.

Click Widening expands the processed time around each detection. Increase it when the initial snap disappears but the sticky tail of a lip smack remains. Reduce it when word attacks soften, pauses sound patched or the module takes a visible bite from neighboring speech.

Nine steps to remove mouth clicks safely

1. Preserve an independent source. Save an untouched audio-file copy and open a separate working copy. A duplicated region or playlist can still point to the same disk file.
2. Select a representative line. Choose the intended file tab, turn off Composite View and Instant Process, and select a phrase with mouth clicks, plosives and crisp consonants.
3. Start conservatively. Open Mouth De-click with low Sensitivity and minimal Click Widening. Leave the Listen ear button off for the first normal-output preview.
4. Preview the phrase. Use the module Preview control and raise Sensitivity only until the distracting clicks are caught. Listen to the surrounding words, not just the isolated click.
5. Audition removed clicks. Activate the Listen ear button and click Preview to hear the removed signal. Lower Sensitivity if wanted consonants or vocal tone dominate the removed signal.
6. Tune the frequency bias. Adjust Frequency Skew for the actual click spectrum. Recheck removed and normal output; do not choose a setting from the speaker’s gender or pitch.
7. Widen only for a tail. Increase Click Widening when a detected smack leaves a short decay. Recheck consonants and return to less widening if speech softens.
8. Return to speech and render. Turn Listen off, preview normal speech again, stop Preview and click Render. Inspect leftovers and spot-repair only the distracting events.
9. Export and approve in context. Keep the editable RX Document if needed, export a new audio filename and reopen it. Compare at matched loudness through the delivery chain, checking duration, alignment, channels and natural diction.

Save a gentle full pass and a separate spot-repaired candidate when both are useful. For each comparison, start from the same source state or undo all experimental renders first; bypassing a module does not undo processing already committed to the file. The RX Undo History lets you return to an earlier state, and an RX Document retains that history.

Follow the RX file-saving rules: Save overwrites an opened WAV or AIFF on disk. Use a distinct export filename for delivery and retain the untreated source. Export File and Export Selection have different scopes; reopen the exported file to confirm that you delivered the intended full recording or selected passage. An RX Document is an editable RX archive, not a DAW audio handoff.

Set Sensitivity without shaving consonants

The most important test is not whether the mouth clicks vanish. It is whether T, K, D, CH and the edges of plosives stay articulate. These transients can resemble the short events the detector is trying to remove.

Loop a line with both the problem and crisp diction. Increase Sensitivity slowly, then activate Listen during Preview in the Editor. A clean removed signal contains ticks, smacks and little else. If recognizable speech rhythm, repeated consonants or vowel brightness appears, the detector is reaching too far.

First compare the untreated and repaired line at matched loudness with later processing bypassed. Then restore the intended chain and listen in the complete mix. Heavy compression or bright EQ can expose remaining clicks; they can also make an over-smoothed consonant easier to miss during a quick solo check. The vendor’s vocal-cleanup walkthrough likewise checks context before detailed repair.

Use Frequency Skew by evidence, not voice type

Frequency Skew is not a “male versus female voice” control. It tells the detector where to emphasize click energy. A low, dull lip separation may need a lower bias; a sharp saliva tick may sit higher. Two clicks from the same speaker can require different spot settings.

In the spectrogram, zoom until the event is visible. A short broadband line suggests an impulse; a thicker mid-frequency patch with a tail suggests a smack. Move Frequency Skew while auditioning the removed signal and stop when the event is captured without collecting excessive vocal formants.

The official iZotope click-removal guide describes Frequency Skew as centering processing on the artifact type. That principle applies here, even though Mouth De-click is the speech-focused detector.

Use Click Widening for tails, not missed detection

Widening does not make the detector smarter. It increases the repair area around events already detected. If the module misses a quiet click entirely, raise Sensitivity carefully or select that event for a separate pass. If it removes the initial click but leaves a short wet decay, increase Click Widening.

Large widening can make a voice sound edited even when no obvious click remains. Listen for rounded consonants, tiny dropouts and unnatural silence between the lips opening and the word beginning. Those are signs that the repair window is wider than the defect.

Use Compare in the Editor to store several candidates, then audition whole sentences instead of clicking repeatedly on the isolated defect. The Preview, Compare and History guide shows how to keep those choices reversible.

Audition removed clicks with Listen

In the RX 12 Editor, the Listen ear button beside Compare lets Preview play the material the module would remove. Use this audition at every major setting change, then turn Listen off to hear processed speech. Older tutorials call the removed-signal monitor Output Clicks Only; the current Editor interface uses the ear button.

Hearing a faint trace of voice is not automatically failure; reconstruction can change a little surrounding energy. Hearing clear words, repeated consonants, pitch or sustained vocal tone is a warning. Reduce Sensitivity, move Frequency Skew closer to the actual click band or shorten widening.

Approve normal speech before Render, then inspect the committed result. Listen auditions removed audio during Preview. Judge the committed repair by playing the processed file, and reopen your export to check the final output. In a host plug-in with an Output Clicks Only control, restore normal output before the bounce. If a committed edit is wrong, undo it or return to the untouched source; bypass alone cannot restore edited samples.

An October 2025 RX 11 support discussion describes a user hearing only clicks after using the monitor. The reply recommends deselecting Output Clicks Only before Render in that older version. This is an individual RX 11 support exchange, not a demonstrated RX 12 rendering defect. Finally listen to normal speech in context: selective removal can still leave an audible patch.

Whole-file pass or spot repair?

Use a whole-file pass when hundreds of similar clicks occur across a podcast, audiobook or voiceover and the conservative setting leaves diction intact. Render only after checking quiet, loud, dry and expressive passages. The worst click is not always the passage most likely to be damaged.

Use spot repair for a few loud smacks, clicks next to consonants or events with different spectra. Select a short region with enough context, preview, process, then move on. This takes longer per event but avoids changing clean speech.

A practical compromise is a gentle global pass followed by marked local repairs. The Mouth De-click manual says a second pass can sometimes catch quieter clicks missed behind louder ones; it does not promise that two passes are always better. Keep the second pass local unless fresh normal-output and removed-signal checks remain acceptable throughout the file.

Where Mouth De-click belongs in a cleanup chain

Repair clicks before heavy EQ, compression and limiting. Compression raises quiet mouth sounds; bright EQ makes high-frequency clicks more obvious. Removing them first gives the later processors less damage to exaggerate.

If clipping is the dominant fault, test De-clip first, then reassess the click detector on that repaired copy. For a combined low-frequency plosive and lip smack, reduce the plosive first and treat the remaining mouth noise separately. Other recordings may need only Mouth De-click. This is a choice by defect, not a mandatory chain applied to every file.

One order constraint is specific: the De-plosive manual recommends using it before a high-pass filter, because filtering can undermine its detection. After local faults, evaluate only the hum, noise, reverb, breaths and tone changes the recording actually needs. A click-free pause is not an improvement if the room tone or intended articulation disappears.

The RX 12 beginner workflow explains selections and versioning, while the background-noise guide separates steady noise from transient mouth events. Do not make one denoiser fight both.

When Mouth De-click is the wrong tool

• Use De-plosive for the low-frequency blast from a consonant.
• Use general De-click for digital ticks, vinyl impulses and waveform discontinuities.
• Use De-crackle for a dense field of small continuous crackles.
• Use Spectral Repair for one irregular sound that can be selected away from speech.
• Use Interpolate for an individual click shorter than 4,000 samples. That is a tiny repair, not a way to rebuild a missing word.
• Re-record when clicks are inseparable from critical consonants and the performance can be recreated.

Mouth De-click is strongest when the defect matches its detector. If the click survives a conservative setting and an aggressive setting damages the word, the artifact needs a different selection or repair model.

Prevent clicks before the edit

Try a slightly off-axis microphone position and a little more distance, then compare a short recording at matched loudness. Keep the script and delivery consistent so the change is meaningful. A pop filter can help with air blasts and spacing, but it is not a cure for saliva clicks. Avoid changing the voice’s tone or adding room sound just to hide one mouth noise.

Have water available and take comfortable breaks during long narration. The iZotope voice-recording guide discusses dry-mouth noise and keeping water nearby. Treat capture changes as things to test, not a guaranteed cure; retain alternate takes and the narrator’s natural performance.

Record matching room tone and retain alternate takes. For a smack entirely in a pause, a careful edit with matching ambience and short crossfades can be preferable to aggressive processing. Do not replace sound under a word with silence: a technically clean gap can damage pronunciation and timing. When an alternate word fits the performance better, use it.

Final Mouth De-click quality gate

• The source and an untreated version remain available.
• Sensitivity catches the problem without removing recognizable speech.
• Frequency Skew follows the click spectrum rather than a voice stereotype.
• Click Widening is used only where a smack has a tail.
• Removed audio and normal speech have both been checked; the committed result contains the intended speech.
• Stubborn leftovers are spot-repaired instead of driving a destructive global pass.
• Plosives, clipping, crackle and broadband noise use their own tools.
• The result is approved at matched loudness in the complete mix.

The Clean Dialogue with RX hub connects Mouth De-click to Dialogue Isolate, De-reverb, De-hum and the rest of a controlled speech-repair workflow.

Frequently asked questions

Which RX 12 editions include Mouth De-click?

RX 12 Standard and Advanced include Mouth De-click as a module and plug-in. RX 12 Elements includes the general De-click plug-in, not the dedicated mouth-noise tool.

What should I set Sensitivity to?

There is no universal number. Start low on a representative line and raise it only until distracting clicks are caught. In the RX 12 Editor, use Listen during Preview to check for wanted speech in the removed signal. Turn Listen off and approve normal speech before Render.

What does Frequency Skew do?

It biases detection toward lower or higher click energy. Set it from the actual artifact and removed signal, not from the speaker's gender or pitch.

When should I increase Click Widening?

Increase it when the initial click disappears but a lip-smack tail remains. It will not detect a missed click and can soften nearby speech when pushed too far.

Should I run Mouth De-click twice?

A second pass can sometimes catch quieter clicks missed behind louder ones. Prefer a conservative first pass and local second-pass selections. Recheck diction after each pass; do not confuse an intentional second pass with accidentally stacking experimental renders.

Why does Mouth De-click damage T and K sounds?

Those consonants contain short transients that can resemble clicks. Lower Sensitivity, reduce widening, adjust Frequency Skew or spot-process only the obvious mouth noises.

Is Mouth De-click better than De-click?

Mouth De-click is dedicated to saliva clicks and lip smacks. General De-click offers broader algorithms, and its Low Latency mode is also documented for mouth clicks. Compare the actual defect; neither is always better on every voice.

Should Mouth De-click come before compression?

Usually yes. Compression can raise quiet mouth sounds and make them harder to ignore, so repair the clicks first and judge the complete chain afterward.