Back to All Events

Preserving DVDs: Agentic AI to the rescue \ Tom Ackroyd

This illustrated talk will describe the successful implementation of an AI-assisted workflow in AV digital preservation.

AV Preservation staff at Ngā Taonga Sound & Vision were recently tasked with preserving a collection of DVD-Video recordings.

At the time we had few established protocols for preservation of DVDs. This issue arose during a project to digitise and preserve a recent nationally significant audiovisual deposit. Resolving this supported our mission to make audiovisual taonga accessible before the original carriers deteriorate beyond rescue.

After reaching consensus on how digital materials derived from DVDs would fit into our established preservation frameworks, the workflow’s design and testing phase hinged on the use of ChatGPT to write a shell script to faithfully extract digital files. Later use of the command-line agent Claude Code extended and formalised both the script and associated GitHub repository. This helped fix many edge-cases and anomalies that arose during the processing of the collection.

Without clear preservation protocols and workflows, audiovisual taonga are at risk. Appropriate collaboration with chatbots and agentic AI in writing code can assist archival institutions where in-house coding experience is limited. This project ensured long-term access to these taonga and established a model for future preservation initiatives.

Tom Ackroyd

With a 40-year background in broadcast television, advertising and video editing, Tom Ackroyd came to Ngā Taonga Sound & Vision in 2016 as a born-digital collector. In 2020 he moved to the Preservation Team where he works to digitise and preserve a broad range of magnetic tape and optical disk audiovisual formats. His favourite format is DVCAM.

Tom lives in Matairangi Mount Victoria in Te Whanganui a Tara Wellington, with his wife Viv and their dog Fred. He is a keen choral singer, photographer and film-goer.

Previous
Previous
24 November

Asking the Shelves: How a District Council Library Built an AI Analyst from 14 Million Transactions \ Angela Jowitt, Andrew Terwiel, Eric Bonhomme

Next
Next
24 November

Our first decade – 10 years of born digital records transfers, reflection and future plans \ Jan Hutař, Carly Lenz