Abstract
The automated generation of long-form audio may present challenges related to vocal consistency and quality control efficiency. A computer-implemented agentic system, which could operate on a server or cloud computing platform, can be used to address such challenges through a recursive review-and-synthesis loop managed by a multi-agent framework. The system may create a multi-dimensional profile for each text segment, defining instructions for attributes such as style, context, and emotion. It can then generate multiple audio candidates from one or more synthesis providers, and an auto-rater agent may score and select candidates based on their alignment with the profile. Low-scoring segments can be recursively re-synthesized with targeted improvement notes. This process may facilitate scalable audio production by enabling iterative self-correction and providing an auditable, segmented review workflow for human-in-the-loop oversight.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Haritaoglu, Ismail and Kuligin, Leonid, "Agentic System for Long-Form Audio Generation Using Recursive Synthesis and Multi-Dimensional Profiling", Technical Disclosure Commons, (May 20, 2026)
https://www.tdcommons.org/dpubs_series/10193