Supplementary code for the paper presented at the DAFx20in22 conference in Vienna.
This repository contains the compile-time and run-time implementations of the temporal convolutional layers and models described in the paper. The implementation itself is inspired by previous work by Damskagg et. al., Steinmetz and Reiss, and the RTNeural library by Jatin Chowdhury. The work aims at minimizing the limitations of previous implementations of Temporal Convolutional Networks for real-time audio signal processing. The main goal was to approach the signal processing speed of Recurrent Neural Networks, which are another suitable option for black-box virtual analog modeling.