A method for immersive audio generation from natural language prompts is presented. We circumvent end-to-end modeling of modern intricate audio formats by utilizing an agentic AI workflow. In addition to a conceptual specifi-cation of the agent, an example immersive generator is implemented. A language model coordination/interface unit is combined with single-source audio models and other tools, and the system is equipped with the proposed theoretical concepts. All communication between the agent, the tools and the user is done via natural language, implying a general concept of language-driven signal processing. The performance is evaluated in listening test with an immersive 7.1.4 loudspeaker layout. The results illustrate the present capabilities of audio agents and give indications of further potential as the individual tools improve.
(See document for exact affiliation information.)
https://aes2.org/publications/elibrary-page/?id=23079
(382KB)
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member Join the AES. If you need to check your member status, login to the Member Portal.
