: This paper details a framework that uses LLMs to decompose high-level user instructions into atomic audio editing steps.