Understanding Sibilance in Restored Recordings

Old audio recordings—whether from analog tape, vinyl, or early digital formats—often carry artifacts that degrade speech clarity. Among the most persistent issues is sibilance, the exaggerated “s,” “sh,” “ch,” and “z” sounds that can pierce through a mix and distract listeners. In restored recordings, sibilance is especially problematic because it is frequently boosted by the noise reduction, equalization, and compression applied during restoration. The result is a brittle, harsh vocal that tires the ear and reduces intelligibility.

Sibilance occurs naturally in human speech, typically in the frequency range of 5 kHz to 10 kHz, with the center of energy varying by speaker and microphone. When a recording is damaged or poorly transferred, these frequencies can become unnaturally prominent. Without careful treatment, a restored recording may sound “clean” on first listen but reveal fatiguing sibilance on repeated playback. De-essing plugins address this by selectively attenuating those frequencies only during the sibilant moments, preserving the natural timbre of the voice during vowels and consonants.

What Is De-essing?

De-essing is a specialized form of dynamic equalization. Unlike a static EQ cut that permanently reduces high frequencies, a de-esser monitors the audio and applies gain reduction only when sibilance exceeds a defined threshold. The plugin detects energy in the sibilant range, compares it to the overall level, and reduces the gain of those frequencies by a ratio you set. This keeps the speech bright and open while taming the offending hiss.

Most de-essers operate in one of two ways: wideband or split-band. Wideband de-essers reduce the overall gain of the entire signal when sibilance is detected, which can cause audible pumping if not tuned carefully. Split-band de-essers (also called multiband compressors with a high-frequency band) only reduce the level of the specific frequency range containing the sibilance, leaving the rest of the spectrum untouched. Many modern plugins combine both approaches and add spectral editing for fine-grained control.

Choosing the Right De-essing Plugin

The de-essing plugin market offers options for every budget and workflow. Below are the primary categories and specific recommendations.

Built-In DAW De-Essers

Almost every major digital audio workstation—including Pro Tools, Logic Pro, Cubase, Ableton Live, and Reaper—includes a stock de-esser. These are often simple to use, with a single frequency knob, threshold, and ratio. While they may lack advanced features like spectrum analyzers or sidechain filters, they are perfectly capable for many restoration jobs. For example, Logic’s De-Esser offers a fast, musical response that works well on dialogue.

Third-Party Plugins (Paid)

  • iZotope RX De-ess: Part of the RX suite, this is the industry standard for post-production and restoration. It includes spectral visualization, a dedicated sibilance detection zone, and options for both wideband and split-band modes. It is especially effective on heavily sibilant recordings because you can see the exact frequencies that peak during “s” and “sh” sounds.
  • FabFilter Pro-DS: Renowned for its transparent processing and adjustable detection range, Pro-DS uses a “Single Vocal Mode” that focuses on eliminating sibilance without affecting breathiness. The sidechain filtering and balance control make it highly adaptable to different source material.
  • Waves DeEsser: A classic, straightforward plugin with dedicated “Male” and “Female” presets. It uses a split-band design and offers a “Split” mode for more aggressive reduction. While not as surgical as newer plugins, it remains a trusty choice for quick results.
  • Sonible de:esser: Uses AI-based detection to adapt to the source material automatically. It offers real-time visualization of the sibilant range and supports both classic de-essing and a “transient” mode for extreme sibilance.

For restoration work, iZotope RX De-ess and FabFilter Pro-DS are particularly recommended because of their precision and low latency. You can learn more about iZotope’s approach on their official site here.

Free and Open-Source Options

  • TDR Nova: A parallel dynamic equalizer that can act as a de-esser. By setting a bell filter around 7 kHz with a moderate Q, enabling dynamic mode, and adjusting threshold and range, you can achieve professional results at no cost.
  • AAMS Auto Audio Mastering Suite: Contains a multiband compressor with a de-essing preset. It is less intuitive than dedicated tools but can work in a pinch.
  • ReaFIR (Reaper): Reaper’s built-in FFT-based dynamic EQ can be configured as a de-esser. While it requires manual setup, it offers incredible flexibility for advanced users.

Step-by-Step De-essing Workflow

Applying a de-esser to a restored recording requires careful listening and iterative adjustment. Follow these steps for best results.

Step 1: Prepare Your Track

Import your restored audio into a DAW or audio editor. Normalize the peak level to around -6 dBFS to provide headroom for processing. If you have already applied noise reduction, broadband EQ, or compression, consider doing a long reverb decay or de-click first, as those processes can alter the sibilant profile.

Step 2: Insert the De-esser and Set the Detection Frequency

Place the de-esser plugin on the track or bus. Listen to a section containing heavy sibilance and solo the plugin’s detection channel (many plugins have a “listen” or “solo” button). Sweep the frequency control to find where the sibilance is most offensive—usually between 5 kHz and 10 kHz. For most male voices, 5–7 kHz is common; for female voices, 7–9 kHz. However, every speaker is different, so trust your ears.

Step 3: Adjust the Threshold

Lower the threshold until you see gain reduction (typically 2–6 dB) on the sibilant consonants. Avoid reducing more than 8–10 dB, as that can cause the speech to sound “lispy” or muffled. The goal is to match the sibilant level to that of the surrounding speech, not to eliminate them entirely.

Step 4: Set Ratio and Attack/Release

A moderate ratio (2:1 to 4:1) works well for restoration. Higher ratios can be used for extreme sibilance but risk artifacts. The attack time should be fast (0.5–2 ms) to catch the transient onset of “s” sounds. Release time should be moderate (10–50 ms) to avoid pumping but fast enough to be ready for the next syllable. Many plugins let you adjust the response curve; start with the “Natural” or “Optimal” preset.

Step 5: A/B and Refine

Toggle the plugin bypass and compare the processed and unprocessed audio. Listen on both studio monitors and headphones. Pay attention to the overall brightness and naturalness. If the speech becomes dull, raise the threshold slightly or reduce the ratio. If sibilance still pokes through, lower the threshold or narrow the detection frequency.

Step 6: Process in Context

If the restored recording includes music or background ambience, engage the de-esser while the entire mix is playing. Sibilance can sound more prominent when other instruments mask the fundamental frequencies of the voice. Adjust the de-esser settings while listening to the full mix to ensure the speech remains clear without over-processing the high frequencies.

Advanced Techniques for Restoration

Sidechain De-essing

Some de-essers (like FabFilter Pro-DS and iZotope RX De-ess) allow you to filter the sidechain signal that triggers the reduction. By high-pass filtering the sidechain above 10 kHz, you can de-ess only the extreme high end while preserving the natural 5–8 kHz region. Alternatively, you can set a band-pass filter around the problem frequency for surgical precision. This technique is especially useful when the sibilance is concentrated in a narrow band.

Manual Automation

If the recording has highly variable sibilance—for example, a speaker who leans closer or farther from the mic—consider automating the threshold or detection frequency over time. In your DAW, draw automation lanes for the de-esser parameters. During loud, sibilant phrases, lower the threshold; during quiet, breathy sections, raise it. This manual approach can yield more transparent results than static settings alone.

Spectral De-essing

Advanced restoration tools like iZotope RX offer a “Spectral De-ess” module that works in the frequency domain. Instead of a single detection band, RX analyzes the entire spectrogram and attenuates only the time-frequency bins that exceed a threshold relative to the surrounding noise floor. This allows for extremely precise removal of sibilance without affecting other high-frequency content like air, teeth clicks, or sibilant “t” and “k” sounds. Use this as a final polish after a traditional de-esser if the sibilance is still uneven.

For a detailed tutorial on spectral de-essing, refer to the iZotope Knowledge Base here.

Parallel De-essing

Parallel processing—mixing the heavily de-essed signal with the original—can preserve natural tonality. Duplicate the track, apply aggressive de-essing to the duplicate, and blend it with the dry track until the sibilance sits comfortably. This technique works best when the de-esser has a “dry/wet” mix knob or you use a send/return routing. Parallel de-essing is especially useful for restoring vintage voiceovers where the original sibilance is part of the character.

Best Practices for Restored Speech

  • De-ess before final limiting or compression: Applying broadband compression after de-essing can reamplify any residual sibilance. Always de-ess earlier in the chain, preferably before any master bus processing.
  • Use the minimal amount of reduction: Over-de-essing makes speech sound unnatural and “scooped.” Aim for subtlety: reduce sibilance until it no longer distracts, not until it disappears.
  • Combine with EQ: A gentle high-shelf cut around 10 kHz after de-essing can tame any residual harshness without affecting the mids. But avoid a static EQ cut at the same frequency as the de-esser, or you risk doubling the reduction.
  • Monitor on multiple systems: Sibilance that sounds fine on studio monitors may become harsh on laptop speakers or earbuds. Check the restoration on consumer playback devices to ensure the de-essing translates.
  • Keep a backup of the raw restoration: If you need to re-evaluate the sibilance later, having an unprocessed version lets you start fresh without cumulative artifacts.
  • Don’t de-ess every track: In a multichannel restoration (e.g., a radio play with multiple voices), only de-ess tracks that exhibit problematic sibilance. Applying it globally can thin out other elements.

For additional reading on audio restoration workflow, Waves has an excellent overview here.

Conclusion

De-essing plugins are indispensable for achieving clear, listenable speech in restored recordings. By understanding the nature of sibilance, selecting the right tool for your needs, and applying a careful workflow—including sidechain filtering, automation, and spectral techniques—you can transform a harsh, fatiguing recording into a polished audio artifact that honors the original performance. The key is to approach de-essing as a surgical task: listen deeply, adjust modestly, and always compare before and after. With practice, you will develop an ear for exactly how much reduction a given recording needs. Whether you are restoring a family tape, a historical broadcast, or a commercial archive, mastering de-essing will dramatically improve the listening experience.

For a deeper dive into advanced restoration techniques, check out this guide from Production Expert here.