AI 'time travel': can artificial intelligence really build you a digital museum?
A De Morgen profile of hobbyists who use AI to restore and reimagine old photographs raises a practical question for anyone in Belgium with a shoebox of family history: how do you 'time travel' responsibly, which Belgian archives can feed the work, and where…
In 30 seconds
- De Morgen profiled hobbyists who use AI to restore, colourise and reconstruct old photographs, with one describing the result as "a kind of digital museum."
- AI image work spans three levels: restoration (fixing damage), colourisation (guessing colour), and generative recreation (inventing new detail).
- Belgian source archives include the State Archives (arch.be), CegeSoma, KIK-IRPA, Amsab-ISG, and city archives such as FelixArchief; Europeana aggregates collections EU-wide.
- Most restoration tools are English-language, while the archives are catalogued in Dutch or French by commune.
The story centres on a growing consumer hobby, profiled by De Morgen, in which people use artificial-intelligence tools to restore, colourise and generatively reconstruct old photographs — a practice one enthusiast called building "a kind of digital museum." Key entities include AI image-restoration and generative models (cloud and open-source), Belgian heritage archives that supply source material — the federal State Archives (Rijksarchief / Archives de l'État), CegeSoma, the Royal Institute for Cultural Heritage (KIK-IRPA), Amsab-ISG, city archives such as Antwerp's FelixArchief, and the EU aggregator Europeana — and the EU Artificial Intelligence Act, which sets transparency rules for AI-generated media.
Background
Photo restoration is as old as photography, but until recently it was skilled manual retouching. Automated colourisation and restoration matured in the late 2010s, and generative models capable of inventing plausible new detail — animating portraits, reconstructing lost scenes — arrived at consumer scale in the 2020s, moving the practice from repair toward reimagining and raising fresh questions about historical authenticity.
What to do
Digitise and keep originals unaltered; use restoration before generative effects; cross-check reconstructions against archive photographs; caption honestly to separate documented history from AI hypothesis; and start the project in the archive, not the software.
Impact
Regional — Flanders is the primary audience touchpoint for this story, and its heritage infrastructure is dense: Amsab-ISG in Ghent, Antwerp's FelixArchief, and many gemeente archives have digitised collections that hobbyists can draw on, catalogued in Dutch and often by historic street name.
EvidenceDeveloping · 1 primary source · Background sources: 2 · some details remain unconfirmedExplore evidence →Hide evidence ↑
- Publication date unavailable
- Retrieved by ODIN:
- 5 Aug 2026
- Publication date unavailable
- Retrieved by ODIN:
- 5 Aug 2026
- Publication date unavailable
- Retrieved by ODIN:
- 5 Aug 2026
- Publication date unavailable
- Retrieved by ODIN:
- 5 Aug 2026
Continue reading
This story was assembled from verified evidence, with its sources and reasoning recorded as it was written.