Ken Axford and Hope Dunbar write: “When faced with over 2,000 hours of historical radio recordings from the former campus station WBFO, the University at Buffalo (N.Y.) Archives team knew that manually creating descriptive metadata would be prohibitively time-intensive. The solution? Leveraging consumer-level artificial intelligence (AI) tools like ChatGPT Pro and Microsoft 365 Copilot to generate program descriptions from audio transcripts. Despite some limitations, the AI-assisted metadata creation significantly reduced processing time for the WBFO collection. A hybrid approach—AI generation followed by human verification—proved both efficient and reliable.”
