PyNeuralFx: A Python Package for Neural Audio Effect Modeling

TL;DR

Neural audio effect modeling suffers from reproducibility and interpretability gaps due to diverse training setups and evaluation criteria. PyNeuralFx offers a standardized Python toolkit with shared model backbones (CNNs and RNNs), conditioning schemes (Concat, FiLM, StaticHyper, DynamicHyper, and hypernetworks), a suite of loss functions, a broad set of objective metrics, and visualization tools. These features enable fair benchmarking, interpretable diagnostics of model behavior (e.g., harmonic response and aliasing), and reproducible experiments via YAML configuration. By unifying methodology and tooling, PyNeuralFx aims to accelerate research in neural audio processing and invites ongoing community contributions and future enhancements.

Abstract

We present PyNeuralFx, an open-source Python toolkit designed for research on neural audio effect modeling. The toolkit provides an intuitive framework and offers a comprehensive suite of features, including standardized implementation of well-established model architectures, loss functions, and easy-to-use visualization tools. As such, it helps promote reproducibility for research on neural audio effect modeling, and enable in-depth performance comparison of different models, offering insight into the behavior and operational characteristics of models through DSP methodology. The toolkit can be found at https://github.com/ytsrt66589/pyneuralfx.

Figures (4)

• Figure 1: Workflow of the proposed PyNeuralFx toolkit.
• Figure 2: Visualization of the harmonic response of a model trained on Boss OD-3 overdrive device yeh24dafx.
• Figure 3: Sine sweep response of a network model, revealing the aliasing problem (i.e., high frequency folds back).
• Figure :