Score Distillation Sampling for Audio: Source Separation, Synthesis, and Beyond

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Richter-Powell, Jessie, Torralba, Antonio, Lorraine, Jonathan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915276315951104
author Richter-Powell, Jessie
Torralba, Antonio
Lorraine, Jonathan
author_facet Richter-Powell, Jessie
Torralba, Antonio
Lorraine, Jonathan
contents We introduce Audio-SDS, a generalization of Score Distillation Sampling (SDS) to text-conditioned audio diffusion models. While SDS was initially designed for text-to-3D generation using image diffusion, its core idea of distilling a powerful generative prior into a separate parametric representation extends to the audio domain. Leveraging a single pretrained model, Audio-SDS enables a broad range of tasks without requiring specialized datasets. In particular, we demonstrate how Audio-SDS can guide physically informed impact sound simulations, calibrate FM-synthesis parameters, and perform prompt-specified source separation. Our findings illustrate the versatility of distillation-based methods across modalities and establish a robust foundation for future work using generative priors in audio tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2505_04621
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Score Distillation Sampling for Audio: Source Separation, Synthesis, and Beyond
Richter-Powell, Jessie
Torralba, Antonio
Lorraine, Jonathan
Sound
Artificial Intelligence
Machine Learning
Multimedia
Audio and Speech Processing
68T07
I.2.6; H.5.5; H.5.1
We introduce Audio-SDS, a generalization of Score Distillation Sampling (SDS) to text-conditioned audio diffusion models. While SDS was initially designed for text-to-3D generation using image diffusion, its core idea of distilling a powerful generative prior into a separate parametric representation extends to the audio domain. Leveraging a single pretrained model, Audio-SDS enables a broad range of tasks without requiring specialized datasets. In particular, we demonstrate how Audio-SDS can guide physically informed impact sound simulations, calibrate FM-synthesis parameters, and perform prompt-specified source separation. Our findings illustrate the versatility of distillation-based methods across modalities and establish a robust foundation for future work using generative priors in audio tasks.
title Score Distillation Sampling for Audio: Source Separation, Synthesis, and Beyond
topic Sound
Artificial Intelligence
Machine Learning
Multimedia
Audio and Speech Processing
68T07
I.2.6; H.5.5; H.5.1
url https://arxiv.org/abs/2505.04621