Signal flow and level controls
Incoming audio drives the cavitation model and is stored as recent grain material. Simulation events read short sections from that history, filter and shape them, then mix the resulting grain cloud with the direct signal.
Input Gain sets the level entering the detector and rolling grain buffer. Raising it makes quieter details excite the simulation more readily; extreme levels can produce intentionally harsh digital clicks and clipping inside dense grain combinations.
Mix moves from direct input to the grain output. At zero the plug-in uses a dry-only bypass and the simulation stops, so it is not merely an output crossfade at that endpoint.
Output Gain sets the final level after the direct and grain signals are combined and before output protection.
The recent-audio buffer, cavitation detection, grain rendering, and particle state all run in the audio processor; the sound does not depend on the interface being open.