Ready to explore! How AI makes heritage easier to discover
1 Jul 2026When searching online for an actor, historic place or sporting event, you only get results if that person, location or milestone has been documented somewhere. And that’s a shame – as it means many heritage stories remain hidden. Over recent years, we’ve been enriching our content partners’ archive material with automatically generated descriptions. Now, we’re developing this further with four new techniques. But what does this make possible, and how do we stay critical? Watch and read on.
The Hermes project is working on several fronts towards one goal: making audiovisual material easier to find and reuse. Our content partners already focus strongly on making their material searchable, but given the volume of data, it’s a hefty task. AI helps us to unearth hidden information on a large scale:
“Who is in the photo?”
We previously applied facial recognition to moving images. In the Hermes project, we’re extending this to photographic material. And we’ve already processed 10,000 photos so far, producing remarkable results…
Do you remember politician Bob Cools? Using facial recognition, we identified his face in videos from later in life. Then, when we applied the same technology to historic glass plates, Bob reappeared - only decades younger! This is how AI is helping us to put names to thousands of anonymous faces.
“What do I hear there?”
A sound recording or video tells more than just what is being said. Sometimes important information lies in applause, laughter or the music – and audio classification allows us to recognise these sounds instantly.
This is particularly useful behind the scenes – since music involves complex rights issues. By automatically spotting these issues, our content partners can more quickly determine which clips can be made available. A handy asset, especially as later this year, thousands of archive clips will be publicly accessible via thearchives.be. With AI’s help, our content partners can identify eligible material far more efficiently.
“Who is speaking?”
Imagine a politician does not appear on screen but speaks during a debate. Clips such as this are currently difficult to find. But voice recognition changes that. Even those off camera can be identified – again improving the discoverability of our partners’ material.
A concrete example? In this interview, Herman De Croo appears to be responding to questions from an unknown interviewer. But voice recognition reveals it’s actually politician Tom Ongena – a name that would otherwise have remained hidden!
“What is this clip about?”
A good title or summary lets you grasp immediately what a clip is about, even without watching it. In previous AI projects, we generated transcripts for our partners’ audio and video material. Now we’re using these to create automatic summaries and titles – valuable information we also hope to share soon on thearchives.be. Stay tuned!
After all, AI is not an end in itself but a tool to make more heritage visible, discoverable and usable. For our content partners, researchers and anyone keen to explore platforms like thearchives.be.
This project is part of the Flemish Government’s Resilience Recovery Plan.
Beelden:
Bob Cools in video: Socialistische Omroep, collection Amsab-ISG, under copyright.
Bob Cools on glass plate: De Volksgazet, collection Amsab-ISG, under copyright.
Interview Herman De Croo: Archive Herman De Croo (archive nr. 1101), collection Liberas, Ghent, CC BY-NC-ND 4.0.