Anonymized Project Name: A Neural Network-Based Granular Synthesizer for Oceanic Soundscapes

Sabina Hyoju Ahn; Ryan Millett; Seyeon Park

Anonymized Project Name: A Neural Network-Based Granular Synthesizer for Oceanic Soundscapes

Abstract:

This paper presents a neural network-based granular synthesizer designed to explore the intersection of sound synthesis, machine learning, and ecological storytelling. The project interprets the environmental transformations caused by the entanglement of synthetic and natural materials, particularly focusing on plastiglomerates—hybrid formations of plastic debris fused with organic elements in marine environments. By organizing granular segments of oceanic field recordings within a latent space, the instrument enables real-time interaction with hybridized sound textures, allowing performers to navigate an evolving sonic landscape. Utilizing autoencoders for dimensionality reduction, spectral analysis, and machine learning-based clustering, the system constructs an abstract space that reflects the hybridization of synthetic and organic materials through sound. The physical instrument integrates gesture-controlled interaction and touch-based interfaces as a means of fostering an intuitive engagement with the sonic footprint of the Anthropocene. Anonymized Project Name contributes to contemporary eco-acoustic and AI-driven sound art by offering a speculative and interactive approach to sonic materiality, inviting reflection on the impact of synthetic waste on natural ecosystems.