Target-specific editions of THE VOID and EverClean for Fourier Audio transform.engine and Waves SuperRack LiveBox. Same audio algorithms, deployment tuned for each server.
These packages keep each plugin's model, coefficients, state transitions and output mathematics intact. The differences are in the Windows execution profile, SIMD dispatch, mono routing, memory behavior and validation tools supplied for the target hardware.
They install beside the standard editions. They are independent Everworks beta builds and are not official Fourier Audio or Waves products.
Compiled for the Intel Core i9-12900 platform and the short processing periods used by transform.engine. Use true mono chains for microphone inputs.
Measured DSP-load reduction: 19 percentage points. Clear the performance history after initialization and complete a long soak test. A recurring Overloaded warning is not acceptable for show use.
The complete THE VOID model and signal path with a dedicated Alder Lake execution profile, AVX-VNNI dispatch when available, and AVX2/SSE2 fallbacks.
The zero-sample EverClean engine with static sequential oneMKL, link-time optimization, preallocated temporal state and an AVX2 live-host path.
Compiled for the Intel Core i7-14700 hardware profile used by SuperRack LiveBox. The packages default to mono so a microphone channel executes one model path.
A separate LiveBox identity with the complete THE VOID model, native VNNI/AVX2/SSE2 dispatch and a contract test for mono SuperRack operation.
EverClean's zero-sample audio algorithm packaged with the dedicated LiveBox compiler profile, static oneMKL and reusable audio-thread state.
Run the included PowerShell installer as Administrator, then rescan VST3 plugins in the host.
For a microphone source, configure 1-in/1-out. Stereo input can execute two model paths and approximately double the channel-dependent work.
Use the physical-host PowerShell validator included in the ZIP. VM measurements are build checks, not server capacity claims.
After all plugins initialize, reset performance data and exercise cues, GUIs and real audio for at least 30 minutes. Do not accept recurring overruns.
