PRESENTATION

Presentation Description:

AI transcription is no longer a future possibility—it’s reshaping how libraries work today.

This co-presented paper from the State Library of New South Wales (SLNSW) and the National Library of Australia explores how two of Australia’s leading cultural institutions are using AI-driven speech-to-text technologies to transform transcription workflows and enhance access to audiovisual collections.

The State Library of NSW has implemented the Amplify platform, combining machine-generated transcripts with community-driven corrections. This model has enabled the Library to make oral history and audio collections more discoverable while engaging the public in meaningful ways. Amplify has proven effective in balancing automation with human insight, particularly for historically rich and culturally significant audio collections. In addition to enhancing access to SLNSW’s own collections, the platform also supports public libraries across New South Wales and other collecting institutions across Australia in making their local oral history collections more accessible.

In contrast, the National Library has taken a different approach. The technology was harnessed to transcribe 57,000 hours of audio in under 3 months, a task that was impossible using traditional transcription processes. Staff focused on developing a tool enabling in-house search across the transcripts and the capacity for human review of the transcripts, with the ability to deliver finalised transcripts via the catalogue. This has unlocked the oral history collection for staff-assisted research queries and other library corporate priorities and will, over time, lead to a transformational change in the number of transcripts that can be publicly accessed. While taking a cautious approach to public access in the short-term, the focus on ensuring corrected transcripts can be accessed via the catalogue will enable enhanced public access in the longer term.
Together, these case studies offer a comparative view of two distinct but complementary approaches to AI transcription in libraries. The paper will explore the benefits and challenges of each model, including considerations around accuracy, accessibility, staff resourcing, community engagement, and long-term sustainability.

We will also reflect on the broader implications of AI transcription for digital access, research, and the evolving role of library professionals. By sharing our experiences, we aim to provide practical insights for other institutions considering similar technologies and to foster a sector-wide conversation about the future of transcription in cultural heritage settings.

This paper highlights that while AI offers powerful tools, thoughtful implementation, collaboration, and human-centred design remain essential to success.