Sound Quality Comparison
Sound example recorded from the ProCo RAT with distortion at noon (0.0), filter at minimum (-1.0) and volume at maximum (1.0). Sound was processed using the best model (in terms of evaluation loss), for each respective configuration, from the longer training run.
Conditioning Noise Examples
Conditioning noise examples for the ProCo RAT models. These examples demonstrate the effect of conditioning noise on the model's output. The first example applies a smooth LFO to the distortion parameter, while the filter and volume are kept constant. The stable models exhibit no control-conditioning induced noise while the unreguralized models produce clicks and oscillations.
This second set of examples demonstrate that even for non smooth conditioning signal (e.g., random uniform values from -0.01 to 0.01), the stable models still exhibit no control-conditioning induced noise.