Synthesizer Preset Interpolation using Transformer Auto-Encoders

TL;DR

This paper presents a transformer-based auto-encoder model for smooth interpolation between sound synthesizer presets, enabling intuitive sound creation and integration into commercial synthesizers.

Contribution

It introduces a bimodal auto-encoder architecture that processes both presets and audio, improving interpolation quality over existing methods.

Findings

The model achieves smoother interpolations than related architectures.

It successfully processes over one hundred FM synthesizer parameters.

The approach can be integrated into commercial synthesizers for real-time use.

Abstract

Sound synthesizers are widespread in modern music production but they increasingly require expert skills to be mastered. This work focuses on interpolation between presets, i.e., sets of values of all sound synthesis parameters, to enable the intuitive creation of new sounds from existing ones. We introduce a bimodal auto-encoder neural network, which simultaneously processes presets using multi-head attention blocks, and audio using convolutions. This model has been tested on a popular frequency modulation synthesizer with more than one hundred parameters. Experiments have compared the model to related architectures and methods, and have demonstrated that it performs smoother interpolations. After training, the proposed model can be integrated into commercial synthesizers for live interpolation or sound design tasks.