Abstract
Descriptive metadata is essential for discovery in audiovisual archives, yet many collections remain minimally described due to limited resources. Beginning in 2024, the University at Buffalo, University Archives launched an ongoing pilot project exploring the use of consumer-level generative AI tools to streamline audio description workflows. Archivists tested ChatGPT and Copilot on transcripts from over 2,000 hours of institutional radio content, using A/B prompt testing to refine outputs and generate concise summaries resembling traditional archival descriptions. The project prioritized free or low-cost AI tools that could be readily adopted by archives with varying resources. To date, 1,230 programs have been described using this approach. While achieving significant efficiency gains, the project also raised ethical questions around privacy, copyright, and professional practice. The pilot demonstrates the potential for AI-assisted description to enhance discovery while highlighting the need for ongoing evaluation of ethical and practical implications in archival work.
Keywords
Get full access to this article
View all access options for this article.