Remove Mouth Clicks in Audacity Using Nyquist

Audacity can automatically remove mouth clicks using specialized Nyquist scripts, offering a free alternative to premium restoration software. While Audacity’s default Click Removal effect struggles with the nuanced, high-frequency transients of saliva noise, custom Nyquist plugins use spectral detection and interpolation to identify and repair these artifacts across entire recordings. This article explains how these Nyquist spectral repair scripts work, how to use them for automated click removal, and their practical limitations.

How Nyquist Scripts Automate Click Removal

Audacity includes a native "Repair" tool, but it requires users to manually select and reconstruct damaged audio in increments of 128 samples or fewer. This manual workflow is impractical for long voiceover or podcast tracks.

To overcome this, developers in the Audacity community created advanced Nyquist scripts, most notably the "De-Clicker" and "De-Esser" plugins developed by Paul Licameli. These scripts automate spectral repair by:

1. Scanning the Frequency Spectrum: The script analyzes the audio for sudden, unnatural spikes in high-frequency energy that characterize wet mouth noises and tongue clicks.
2. Isolating the Transients: It separates the brief transient spike from the underlying sustained vocal tone.
3. Spectral Interpolation: Similar to manual spectral editing, the algorithm reconstructs the missing or damaged frequencies using the spectral data immediately preceding and following the click.

How to Use the Nyquist De-Clicker Plugin

To automatically remove mouth clicks across an entire track:

1. Obtain the Plugin: Download the DeClicker.ny script from the official Audacity forum's Nyquist plugin repository.
2. Install the Script: In Audacity, go to Tools > Plugin Manager, click Add/Remove Plugins, select the file, and click Enable.
3. Select Audio: Highlight your vocal track or a specific section containing mouth noise.
4. Configure Settings: Open the effect from the Effect menu. For speech, adjust the threshold and frequency limits:
   • Sensitivity: Controls how aggressively the script flags potential clicks. Higher settings catch quieter clicks but increase processing time.
   • Min/Max Frequency: Restrict detection to the typical mouth-click zone (often between 2,000 Hz and 12,000 Hz) to avoid mistaking plosives or consonants for clicks.
5. Process: Run the effect. The script will automatically parse the file and repair detected clicks sequentially.

Performance and Practical Limitations

While Nyquist spectral scripts provide automated repair, they operate within certain boundaries:

• Processing Time: Nyquist scripts run on an interpreted language inside Audacity, making them significantly slower than native C++ effects or VST plugins. Processing a one-hour podcast can take several minutes depending on CPU performance.
• False Positives: Aggressive settings can flag crisp consonants—such as "t," "k," and "c"—as mouth clicks, potentially softening speech diction. Always test settings on a short sample before applying them to a full project.
• Severe Clicks: Extremely loud or long clicks may exceed the script's repair duration limits, requiring targeted manual correction using Audacity's Spectral Selection and Multi-Tool.