An Octave-based Multi-Resolution CQT Architecture for Diffusion-based Audio Generation
Abstract
This paper introduces MR-CQTdiff, a novel neural-network architecture for diffusion-based audio generation that leverages a multi-resolution Constant- Transform (CT). The proposed architecture employs an efficient, invertible CQT framework that adjusts the time-frequency resolution on an octave-by-octave basis. This design addresses the issue of low temporal resolution at lower frequencies, enabling more flexible and expressive audio generation. We conduct an evaluation using the Fr\'echet Audio Distance (FAD) metric across various architectures and two datasets. Experimental results demonstrate that MR-CQTdiff achieves state-of-the-art audio quality, outperforming competing architectures.
Keywords
Cite
@article{arxiv.2509.16603,
title = {An Octave-based Multi-Resolution CQT Architecture for Diffusion-based Audio Generation},
author = {Maurício do V. M. da Costa and Eloi Moliner},
journal= {arXiv preprint arXiv:2509.16603},
year = {2025}
}
Comments
accepted at IEEE International Symposium on the Internet of Sounds