Evaluating Latent Space Structure in Timbre VAEs: A Comparative Study of Unsupervised, Descriptor-Conditioned, and Perceptual Feature-Conditioned Models

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Cameron, Joseph, Blackwell, Alan
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912971165270016
author Cameron, Joseph
Blackwell, Alan
author_facet Cameron, Joseph
Blackwell, Alan
contents We present a comparative evaluation of latent space organization in three Variational Autoencoders (VAEs) for musical timbre generation: an unsupervised VAE, a descriptor-conditioned VAE, and a VAE conditioned on continuous perceptual features from the AudioCommons timbral models. Using a curated dataset of electric guitar sounds labeled with 19 semantic descriptors across four intensity levels, we assess each model's latent structure with a suite of clustering and interpretability metrics. These include silhouette scores, timbre descriptor compactness, pitch-conditional separation, trajectory linearity, and cross-pitch consistency. Our findings show that conditioning on perceptual features yields a more compact, discriminative, and pitch-invariant latent space, outperforming both the unsupervised and discrete descriptor-conditioned models. This work highlights the limitations of one-hot semantic conditioning and provides methodological tools for evaluating timbre latent spaces, contributing to the development of more controllable and interpretable generative audio models.
format Preprint
id arxiv_https___arxiv_org_abs_2603_16713
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Evaluating Latent Space Structure in Timbre VAEs: A Comparative Study of Unsupervised, Descriptor-Conditioned, and Perceptual Feature-Conditioned Models
Cameron, Joseph
Blackwell, Alan
Sound
We present a comparative evaluation of latent space organization in three Variational Autoencoders (VAEs) for musical timbre generation: an unsupervised VAE, a descriptor-conditioned VAE, and a VAE conditioned on continuous perceptual features from the AudioCommons timbral models. Using a curated dataset of electric guitar sounds labeled with 19 semantic descriptors across four intensity levels, we assess each model's latent structure with a suite of clustering and interpretability metrics. These include silhouette scores, timbre descriptor compactness, pitch-conditional separation, trajectory linearity, and cross-pitch consistency. Our findings show that conditioning on perceptual features yields a more compact, discriminative, and pitch-invariant latent space, outperforming both the unsupervised and discrete descriptor-conditioned models. This work highlights the limitations of one-hot semantic conditioning and provides methodological tools for evaluating timbre latent spaces, contributing to the development of more controllable and interpretable generative audio models.
title Evaluating Latent Space Structure in Timbre VAEs: A Comparative Study of Unsupervised, Descriptor-Conditioned, and Perceptual Feature-Conditioned Models
topic Sound
url https://arxiv.org/abs/2603.16713