Discover how Sybilla, our new de-esser, works internally
HoRNet Sybilla: Intelligent, Minimalist De-Essing
Sybilla was created with a single purpose: to remove sibilance from audio tracks, especially vocals. The goal was to design a highly usable plugin that aligns with the HoRNet philosophy—use the least possible number of controls while embedding as much intelligence as possible into the internal processing. The result is a tool that allows audio engineers to work faster, spending less time tweaking and more time creating.
The development of Sybilla began with a simple question: has anyone already solved the problem of de-essing in an elegant, analog fashion? During this exploration, the legendary DBX 902 stood out. Its solution was remarkably efficient. Rather than providing a standard set of de-essing controls, it featured only two main parameters: a frequency selector and an effect amount knob. Notably, the DBX 902 removed the threshold control altogether. The unit automatically calculated the threshold relative to the input signal level, making the de-essing process effective across both low and high signal levels. There was no need for constant threshold adjustments.
Inspired by this, Sybilla adopts the same core principle but takes it a step further through digital processing. The incoming signal is split into two paths: the original audio and a sidechain path. The sidechain is further divided into six distinct frequency bands, each centered around the range where sibilance commonly occurs—typically between 3kHz and 10kHz. These six bands, along with the original signal, create a total of seven signals that are analyzed in real time.
Each signal is passed through an envelope follower, which reacts at a speed similar to that of a VU meter—an intentional choice since VU meters were originally calibrated to reflect vocal intensity. These envelope followers measure the energy within each band. Once the signal levels are extracted, a comparison process takes place: the intensity of each of the six bands is compared to that of the full-band original signal.
This is where the magic of Sybilla happens. If any of the six frequency bands exceeds the intensity of the original signal, that band is flagged as potential sibilance. It is then used to trigger a VCA compressor, which performs the actual gain reduction for de-essing. This intelligent method allows Sybilla to apply targeted compression only where needed, avoiding the artifacts and dullness that can occur when broader de-essing techniques are applied across an entire frequency range.
To give the user control over how aggressive the de-essing is, Sybilla includes an “Effect Amount” knob. Rather than acting as a traditional mix or output gain control, this knob adjusts the compression ratio. For the same difference between the sidechain and the main signal, a higher effect amount applies stronger gain reduction. This makes the de-essing scalable, giving engineers the freedom to adapt the plugin’s behavior to different voices, styles, or mixing needs.
To complement this flexibility, Sybilla also includes a Dry/Wet control. This allows the user to blend the original signal back in with the processed output. It’s particularly useful in situations where hard compression might remove too much high-frequency content, causing the vocal to lose brilliance or presence. With the dry/wet mix, you can strike the perfect balance between clean de-essing and natural tone preservation.
The overall user interface of Sybilla remains faithful to the minimalism that inspired its creation. No unnecessary meters or graphs distract the user. Instead, you’re given just enough control to adjust the core behavior of the plugin, while the internal algorithms do the heavy lifting. This design enables Sybilla to function equally well for seasoned engineers looking to streamline their workflow and beginners who need reliable results without complex setups.
In summary, Sybilla offers a modern, intelligent approach to de-essing inspired by proven analog concepts but enhanced through digital precision. Its internal multi-band detection, envelope-following dynamics, and context-aware compression deliver professional-quality results with minimal user intervention. Whether you’re treating a harsh vocal in a pop track, cleaning up spoken word in a podcast, or smoothing out high-frequency spikes in a live recording, Sybilla adapts to the task with ease and accuracy.
As always, feedback is welcome. If you have any suggestions or observations about how Sybilla works for you, we encourage you to share them. This plugin is the result of thoughtful design and experimentation, but the real value comes from how it fits into your workflow and supports your creativity.
