SAO-Instruct: Free-form Audio Editing using Natural Language Instructions

Michael Ungersböck, Florian Grötschla, Luca A. Lanzendörfer, June Young Yi, Changho Choi, Roger Wattenhofer · 2025

Read SAO-Instruct: Free-form Audio Editing using Natural Language Instructions by Michael Ungersböck, Florian Grötschla, Luca A. Lanzendörfer, June Young Yi, Changho Choi, Roger Wattenhofer (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-22795.txt.

Description

Editing audio with plain words has lagged behind audio creation. Michael Ungersbock leads a team with five co-authors. In 2025 they introduce SAO-Instruct. It edits audio clips from any free form instruction. Older tools need a full description of the result. Others allow only fixed edit types.

The model builds on Stable Audio Open. Training data are triplets of input, instruction, and output. Those triplets come from prompt tools and manual edits. Some training data are synthetic. Still, the model handles real wild clips well. It also handles unseen instructions. Listening tests favor it over rival editing methods. Code and weights are released.

Catalog record

CREATORMichael Ungersböck, Florian Grötschla, Luca A. Lanzendörfer, June Young Yi, Changho Choi, Roger Wattenhofer
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE61 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-22795

Browse the full library · About the Tower of Babel