Lightweight and interpretable neural modeling of an audio distortion effect using hyperconditioned differentiable biquads

TL;DR

This paper introduces a lightweight, interpretable neural model for audio distortion effects using hyperconditioned differentiable biquads, enabling efficient, stable, and understandable modeling of analog audio circuits.

Contribution

The work extends trainable IIR filters to hyperconditioned models with a novel Fourier-based training scheme, resulting in fewer parameters and enhanced interpretability compared to traditional methods.

Findings

Successfully models a BOSS MT-2 pedal at 44.1 kHz

Uses only 40 biquads and 210 parameters

Models are interpretable and can be adjusted by non-experts

Abstract

In this work, we propose using differentiable cascaded biquads to model an audio distortion effect. We extend trainable infinite impulse response (IIR) filters to the hyperconditioned case, in which a transformation is learned to directly map external parameters of the distortion effect to its internal filter and gain parameters, along with activations necessary to ensure filter stability. We propose a novel, efficient training scheme of IIR filters by means of a Fourier transform. Our models have significantly fewer parameters and reduced complexity relative to more traditional black-box neural audio effect modeling methodologies using finite impulse response filters. Our smallest, best-performing model adequately models a BOSS MT-2 pedal at 44.1 kHz, using a total of 40 biquads and only 210 parameters. Its model parameters are interpretable, can be related back to the original analog audio circuit, and can even be intuitively altered by machine learning non-specialists after model training. Quantitative and qualitative results illustrate the effectiveness of the proposed method.