Mastering Mouth Sounds and Breath Removal in Dialogue

Every dialogue editor quickly discovers that the most subtle sounds—a quiet lip smack, a sharp click, an audible inhale—can derail an otherwise polished recording. While audiences may not consciously notice these artifacts, their cumulative effect is listening fatigue and reduced immersion. Removing them without creating robotic or hollow dialogue requires a blend of technical skill, good tools, and an ear for natural dynamics. This article explores advanced techniques that go beyond simple cuts and fades, offering methods to clean dialogue while preserving its life and realism.

Anatomy of Unwanted Vocal Artifacts

Mouth Sounds: Clicks, Pops, and Lip Smacks

Mouth sounds occur when saliva, tongue, or lips create sudden transient noises. Clicks and pops often appear in the waveform as sharp, narrow spikes, typically in the 2–8 kHz range. Lip smacks are lower-frequency, broader events around 200–500 Hz. Recognizing these spectral signatures is critical because blunt removal tools can accidentally erode the sibilance or vowel clarity of the speech itself. A click that lands right on the attack of a “t” or “k” sound can be mistaken for part of the consonant, so careful listening is essential.

Breaths and Their Role in Performance

Breaths are natural—they signal emotion, pacing, and realism. Over-removal leads to the “plastic” dialogue effect that plagues amateur productions. The goal is not silence but control; an unobtrusive breath that sounds human but doesn't distract. Advanced editors preserve softer breaths while attenuating those that are harsh, loud, or placed awkwardly (e.g., in the middle of a phrase). Understanding the actor’s delivery helps you decide which breaths to keep and which to tame.

Pre-Editing Preparation

Before applying any removal technique, listen to the entire dialogue clip at a consistent monitoring level. Note patterns: does the speaker have chronic dry-mouth clicks at phrase starts? Are there inhalations that sound like wind noise? This analysis helps you choose the right tool for each type of artifact. Always work on a copy of the raw file, and keep the original as a safety net. A good practice is to create a session with multiple playlists or versions so you can compare edit decisions later.

Advanced Spectral Editing

Visual Identification in a Spectrogram

Most professional digital audio workstations (DAWs) and standalone editors—such as iZotope RX, Acon Digital Extract Dialogue, or Adobe Audition—offer a spectral display. Mouth clicks appear as short, vertical lines with a bright energy peak. Breaths often show a downward sloping energy from low to mid frequencies, with no harmonic structure. By selecting only these small regions, you can attenuate or remove them without touching the surrounding voice. Practice identifying these shapes across different speakers; each person has a unique spectral fingerprint for their mouth noises.

Attenuation vs. Silent Removal

For clicks, a 6–12 dB reduction over a few milliseconds often suffices rather than a full cut. Breaths can be lowered by 3–6 dB using a volume envelope or a spectral gain brush. The key is to use gentle, precise edits: select only the artifact’s footprint in the spectrogram, then apply a subtle gain reduction. Over-zealous removal leaves a flat, unnatural silence that the ear detects as a missing moment. When you use spectral attenuation, always listen to the result in the context of the full mix before committing.

Repairing Click Artifacts with De‑Click Modules

Many advanced tools include a dedicated “De‑click” or “Mouth De‑click” module. These analyze the audio for transients that match click signatures and apply a multi‑band reduction. The best practice is to set the sensitivity low enough to catch clicks but not to treat sibilant “s” sounds as clicks. Listen to the result in both solo and full mix contexts, and adjust the filter strength accordingly. Some modules also allow you to view a “learned” noise profile, which can improve accuracy for recurring mouth sounds.

Dynamic Noise Reduction and Adaptive Filters

Real‑Time vs. Offline Processing

Dynamic noise reduction (DNR) plugins like Waves WLM or FabFilter Pro‑DS can be applied in real time, but they are less precise than offline spectral editors. DNR works best for constant background noise (hum, hiss) rather than transient mouth clicks. For breaths, DNR can be set to attenuate low‑level noise below a threshold, effectively softening breaths that fall below the speech level. However, it may also soften quiet speech syllables. Always use a sidechain or frequency‑dependent threshold to protect vocal intelligibility. Offline processing gives you the ability to undo and fine‑tune, while real‑time processing is better for quick previews during a mix.

Adaptive Threshold Gating

A modern approach uses an adaptive noise gate that responds to spectral changes. For example, you can set a gate to close only when the signal matches a noise profile of a typical breath (low energy at high frequencies, high energy at low frequencies). This allows breaths to be attenuated without affecting speech. Tools like iZotope RX Breath Control or Accusonus ERA Noise Remover specialize in this. They often include a “tolerance” parameter that lets you decide how much of the breath shape to preserve. Start with conservative settings, then fine‑tune based on the speaker’s dynamic range.

Manual Editing Workflow for Precision

Zoom and Identify

Set your DAW waveform view to show details down to the sample level. A clear mouth click will appear as a sharp spike. Listen to the event in isolation (use a short region around the artifact). If you’re confident it’s unwanted, you have several options:

  • Gain reduction: Draw an automation lane to pull down the volume by 6–12 dB for 10–30 ms.
  • Crossfade removal: Cut the click and apply a very short crossfade (1–3 ms) between the surrounding audio. This works best for single‑sample clicks.
  • Spectrum replacement: In spectral editors, you can “paint over” a click with adjacent audio information. Useful for clicks that occur inside a vowel where cutting would cause a glitch.

Breath Editing with Fades

For breaths, use fade‑in/fade‑out curves to smooth the transition. If you need to remove a breath entirely, cut between the last word and the breath, then crossfade over 10–20 ms to avoid a pop. Alternatively, reduce the breath’s volume by 50–70% using an automation clip. Always listen to the phrase twice: once at normal level, once boosted to check for audible artifacts. For breaths that feel too long, you can trim the tail end to tighten the pace without making it sound chopped.

Tool‑Specific Recommendations and External Resources

While the techniques are universal, certain tools excel at mouth sound removal. The following external resources provide deeper training and practical examples:

  1. iZotope’s official guide on removing mouth clicks with RX – a thorough walkthrough of spectral editing and the De‑click module.
  2. Sound On Sound: Dialogue Editing – Cleaning Breaths – a classic article covering manual breath control techniques and the philosophy behind natural editing.
  3. Produce Like a Pro: Removing Mouth Sounds – Advanced Tips – practical advice from professional audio engineers, including microphone placement insights that reduce mouth noise at the source.
  4. Video Tutorial: Dialogue Cleanup Workflow – a step‑by‑step visual guide from a post‑production specialist, showing a real session.

Advanced Workflows Combining Multiple Techniques

No single technique works for every recording. A production‑ready workflow often combines spectral de‑click for transient clicks, dynamic breath reduction for moderate breaths, and manual edits for stubborn artifacts. Here is a suggested order that many professional editors follow:

  1. Apply a gentle spectral de‑click (threshold around 6–8) to catch the obvious clicks without affecting sibilants.
  2. Use a breath attenuation tool (e.g., RX Breath Control) set to -4 dB reduction on breaths longer than 300 ms.
  3. Listen through the entire dialogue, marking any remaining artifacts (lip smacks, heavy breaths).
  4. Manually edit those with volume automation or spectral gain. For lip smacks, a 3–6 dB reduction in the 200–500 Hz range often works.
  5. Check the final mix at different playback levels. Sometimes a breath that sounds loud in solo is perfectly natural in a full mix with music and effects.

Adjust the order based on the severity of the issues. If a recording has excessive lip smacks, you might want to manually handle those before running a de‑click module to avoid false positives.

Maintaining Natural Dynamics: The Art of “Invisible” Editing

The Rule of Context

Breaths can convey emotion: a quick gasp shows surprise, a slow exhale suggests relief. Stripping all breaths removes these cues. Instead, shape breaths to support the performance. Roll off the low rumble of a breath (below 100 Hz) to reduce bass boom while leaving the airy texture. For mouth clicks, consider whether they might be part of a character’s vocal character (e.g., a dry, older voice). Only remove clicks that are repetitive or loud enough to distract. In narrative contexts, a few retained clicks can actually enhance the sense of realism.

Volume Consistency Without Compression

A common mistake is to rely on heavy compression to even out breaths. Instead, use volume automation to manually adjust breath levels relative to surrounding speech. This preserves the original dynamic range and avoids pumping artifacts. Use a 5–10 ms attack and 20–50 ms release for any gentle compression you do apply, focusing only on the 200 Hz–4 kHz range where speech intelligibility lives. Multiband compression can also help by targeting the low‑mid frequencies that amplify breath rumble without squashing the vocal clarity.

Troubleshooting Common Problems

Problem Likely Cause Solution
Dialogue sounds hollow after click removal Overlapping removal with sibilance or vowel harmonics Use narrower spectral selection; attenuate by 3–6 dB, not full silence
Breaths sound suddenly cut off Too aggressive fade‑out or gate threshold Extend fade‑out to 50–100 ms; lower threshold to allow softer breaths
Lip smacks still audible after de‑click Frequency range too narrow; smacks are low‑frequency Add a separate de‑pop or low‑frequency attenuation step (200–500 Hz)
Mouth clicks in middle of words Misidentification of plosive or sibilant Listen to the event in isolation; if it’s part of the word (e.g., “t” release), leave it
Breath reduction makes dialogue sound “breathy” Too much high‑frequency attenuation on breaths Use a low‑shelf filter instead of full cut; preserve the air above 4 kHz

Ethical Considerations in Dialogue Editing

Dialogue editing is a restoration process, not a creative rewrite. Removing every breath or click can cross into unnatural territory. In narrative audio, the audience expects a degree of human imperfection. Over‑editing can diminish the emotional impact of a performance. Always ask: “Does this edit serve the story and the character?” If the answer is uncertain, err on the side of preserving the original performance. A slightly clicky but emotional take is often better than a sterile, lifeless one. For documentary or reality content, over‑cleaning can make the subject sound manufactured, reducing authenticity.

Building Your Personal Editing Workflow

Experimentation is key. Set aside a few test takes and try different combinations of spectral editing, dynamic reduction, and manual adjustments. Note which techniques work best for different microphone types (e.g., close‑mic’d voice‑overs vs. boom‑recorded dialogue). The proximity effect of a close mic often exaggerates low‑frequency mouth noises, while a boom may pick up more ambient breath noise. Document your settings and results so you can recall them for future projects. Over time, you will develop a workflow that is both efficient and respectful of the source material.

Final tip: Listen to your edited dialogue on multiple playback systems—headphones, nearfield monitors, and a consumer device like a phone speaker. What sounds pristine in the studio may reveal unnatural edits in a less forgiving environment. Adjust accordingly, especially for clicks that become more apparent on small speakers due to frequency masking.

By mastering these advanced techniques, you can produce dialogue that is clean, clear, and still full of the natural nuances that make spoken communication compelling. The best editing is the kind the audience never notices. Trust your ears, respect the performance, and keep refining your process with every project.