Taming Audio VAEs via Target-KL Regularization

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Seetharaman, Prem, Kumar, Rithesh
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866917504977207296
author Seetharaman, Prem
Kumar, Rithesh
author_facet Seetharaman, Prem
Kumar, Rithesh
contents Latent diffusion models have emerged as the dominant paradigm for many generation tasks including audio generation such as text-to-audio, text-to-music and text-to-speech. A key component of latent diffusion is an autoencoder (VAE) that compresses high-dimensional signals into a low frame rate continuous representation that is conducive for downstream prediction. Regularizing these VAEs is challenging, as there is a trade-off between over-regularized (poor output quality) and under-regularized (difficult to predict) latent representations. We propose a framework for studying this trade-off through compression and train Audio VAEs at specific bitrates via target-KL regularization. This allows direct comparison to well-studied discrete neural audio codec models, and the construction of rate-distortion curves for audio VAEs. We evaluate the impact of target-KL regularization on text-to-sound generation and find that sweeping compression rates is helpful in identifying the optimal generation setting.
format Preprint
id arxiv_https___arxiv_org_abs_2605_17085
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Taming Audio VAEs via Target-KL Regularization
Seetharaman, Prem
Kumar, Rithesh
Sound
Machine Learning
Audio and Speech Processing
I.2.7; H.5.5
Latent diffusion models have emerged as the dominant paradigm for many generation tasks including audio generation such as text-to-audio, text-to-music and text-to-speech. A key component of latent diffusion is an autoencoder (VAE) that compresses high-dimensional signals into a low frame rate continuous representation that is conducive for downstream prediction. Regularizing these VAEs is challenging, as there is a trade-off between over-regularized (poor output quality) and under-regularized (difficult to predict) latent representations. We propose a framework for studying this trade-off through compression and train Audio VAEs at specific bitrates via target-KL regularization. This allows direct comparison to well-studied discrete neural audio codec models, and the construction of rate-distortion curves for audio VAEs. We evaluate the impact of target-KL regularization on text-to-sound generation and find that sweeping compression rates is helpful in identifying the optimal generation setting.
title Taming Audio VAEs via Target-KL Regularization
topic Sound
Machine Learning
Audio and Speech Processing
I.2.7; H.5.5
url https://arxiv.org/abs/2605.17085