Real-Time De-Essing in Ecasound via Inverted Bandpass

Ecasound can perform real-time de-essing on vocal tracks by utilizing parallel signal chains, bandpass filtering, and phase inversion. By splitting an incoming vocal signal into a dry chain and a dynamic sidechain, you can isolate harsh sibilant frequencies, invert their polarity, and dynamically blend them back into the main mix to cancel out excessive high-frequency energy. This technique creates a functional, dynamic notch filter that reacts in real time to vocal sibilance using low-latency command-line processing.

The Mechanics of Phase-Inversion De-Essing

Traditional de-essing relies on a dynamic equalizer or a sidechain compressor keyed to high frequencies. In Ecasound, where native dynamic sidechain routing can be complex, phase cancellation offers an effective alternative:

1. Dry Path: The vocal track passes through unprocessed or with general channel equalization.
2. Sibilance Path: The identical vocal signal passes through a steep bandpass filter centered on the problem frequencies (typically 4 kHz to 8 kHz).
3. Phase Inversion: The isolated sibilance is phase-inverted (multiplied by a negative gain value).
4. Dynamic Attenuation: When the phase-inverted sibilant signal is summed with the dry path, the targeted frequencies cancel out. By gating or dynamically controlling the amplitude of this inverted signal based on vocal intensity, cancellation only occurs when harsh sibilance crosses a specific threshold.

Configuring the Chains in Ecasound

To build this in Ecasound, assign the input to two distinct chains within a single chainsetup:

ecasound -B:rt -b:256 \
  -a:dry,sibilance -i:jack_auto,system:capture_1 \
  -a:dry -o:jack_auto,system:playback_1 \
  -a:sibilance -efb:6000,1500 -ea:-100 -o:jack_auto,system:playback_1

In this routing:

• -a:dry,sibilance -i:... routes the real-time input to two parallel chains simultaneously.
• -a:dry routes the unaffected audio directly to the output.
• -a:sibilance -efb:6000,1500 isolates frequencies around 6 kHz with a 1.5 kHz bandwidth.
• -ea:-100 inverts the phase of the filtered signal by applying a negative gain modifier.

Implementing Dynamic Control

A static inverted bandpass simply acts as a standard static notch filter. To make the de-esser dynamic, the inverted chain must remain silent until a sibilant peak occurs.

In pure Ecasound environments, this dynamic action is achieved by inserting a downward expander or noise gate on the inverted chain before the phase-inverting gain stage. When the vocalist sings vowels, the gate stays closed, leaving the dry signal untouched. When high-frequency sibilants occur, the energy opens the gate, introducing the inverted waveform to subtract the harshness.

For more precise dynamics, load LADSPA plugins directly inside the Ecasound chainsetup. Using a dedicated LADSPA compressor/limiter on the bandpass chain allows you to shape the attack and release times of the cancellation, ensuring smooth transitions without phase artifacts or comb filtering on unvoiced consonants.