Release Notes:
- Threshold knob and detection meter slider: They now control the Teager-Kaiser Energy Operator, basically how sharp or erratic an energy spike needs to be before the algorithm classifies it as sibilance rather than part of the vocal cord's natural vibration.
- Reprogrammed Absolute Mode: It learns the singer’s actual voice in real time, If a sound doesn't align with that "learned subspace" (like a sharp burst of air), the Orthogonal energy gating removes it. In absolute mode the de-esser uses 3-vector space to perform the separation
- Voiced Axis (Harmonics): Where the periodic, "vowel-like" energy lives.
- Unvoiced Axis (Sibilance): Where the aperiodic, TKEO-detected "noise" lives.
- Residual/Error Axis: The "math dust" that doesn't fit either category.
- Reprogrammed Relative Mode: Just like absolute mode it too learns the singer’s actual voice in real time, If a sound doesn't align with that "learned subspace" (like a sharp burst of air), the Orthogonal energy gating removes it. However, it switches the plugin to operate in Multi-Vector Space (N-dimensions) for more complex separation:
- Higher-Order Correlation: It’s not just looking at Harmonics vs. Noise. It starts looking at secondary relationships—like how the air at 12kHz specifically correlates with the chest resonance at 300Hz.
- Contextual Intelligence: It decides which extra vectors to add based on the complexity of the signal. If you have a breathy singer with complex "stacking" textures, the math expands to map those unique characteristics so it doesn't accidentally categorize a "cool breath" as a "bad sibilant."
- Subspace Contextuality: allows the subspace to expand and contract its dimensions in real-time, making it significantly more transparent for singers with a lot of dynamic "character" or those who shift between airy whispers and belts.
- Reprogrammed Split/Wide switching: It now controls the following parameters:
- Split = Harsh-band analysis: The detector focuses mostly on the sibilant region set by the frequency selector, and compares it to the whole signal to determine how it should be processed.
- Wide = Full-signal analysis: The detector looks at the entire signal, not just the selected region, it actively detects sibilance across the whole signal and processes the whole signal accordingly
Note:
For users new to CLAP plugins, they can sometimes look like folders on macOS, but the name of the folder has ".clap" in it like a file extension. It's perfectly normal.
Note:
The zip files contain both CLAP and VST3 plugins.
Note for macOS users:
macOS Gatekeeper blocks the binary because it has no code signature. Locally-built binaries are trusted automatically; externally built ones are flagged as "from the internet".
-
To fix this problem after unzipping run the following command:
xattr -dr com.apple.quarantine [path of the Nebula DeEsser.clap or Nebula DeEssr.vst3 file] -
After that you can copy it to either /Library/Audio/Plug-Ins/CLAP or /Library/Audio/Plug-Ins/VST3 (if you want to install it for all users) or ~/Library/Audio/Plug-Ins/CLAP/ or ~/Library/Audio/Plug-Ins/VST3/ (if you want to install it for only the current user)