Trend AnalysisArts & Design

Digital Preservation of Cultural Heritage: AI, GANs, and the Race Against Time

Cultural heritage is disappearing faster than humans can document it. AI-powered tools—from GANs that reconstruct lost artifacts to Heritage BIM that digitizes entire historical sites—are transforming preservation from reactive rescue to proactive digital archiving.

By Sean K.S. Shin
This blog summarizes research trends based on published paper abstracts. Specific numbers or findings may contain inaccuracies. For scholarly rigor, always consult the original papers cited in each post.

Why It Matters

UNESCO estimates that intangible cultural heritage—oral traditions, performing arts, social practices, and traditional craftsmanship—is disappearing at an accelerating rate due to urbanization, globalization, and environmental degradation. Physical heritage faces parallel threats: conflict, climate change, and neglect destroy irreplaceable sites and artifacts every year. Traditional documentation methods (photography, written records, physical conservation) cannot keep pace with the scale of loss.

Digital preservation technologies offer a fundamentally different approach: creating comprehensive digital twins of cultural artifacts and sites that can survive the destruction of their physical originals. Recent advances in generative AI, 3D scanning, and Heritage Building Information Modelling (HBIM) have made it possible to reconstruct damaged or lost heritage with unprecedented fidelity—raising both exciting possibilities and difficult questions about authenticity.

The Science / The Practice

Multimodal AI for Intangible Heritage

Li (2025) presents a multimodal framework integrating Generative Adversarial Networks (GANs) and Natural Language Processing (NLP) specifically for intangible cultural heritage preservation. The framework addresses a problem that physical digitization cannot solve: intangible heritage exists as practices, knowledge, and skills rather than objects. Li's approach uses CNNs for image recognition of craft techniques, GANs for generating visual representations of endangered practices, and Transformer-based NLP for processing oral tradition narratives. The application to rural revitalization is particularly significant—connecting digital preservation to economic sustainability for communities that are the living carriers of heritage.

Text-to-Image Reconstruction

Hsieh et al. (2024), with 2 citations, explore how advanced text-to-image models—specifically CLIP adapted for Chinese cultural contexts—can reconstruct cultural heritage from textual descriptions. This addresses a common scenario in heritage preservation: many historical artifacts exist only in written records, with no surviving visual documentation. The ability to generate plausible visual reconstructions from historical texts opens new possibilities for museums, educators, and researchers, though the authors carefully note the distinction between AI-generated reconstructions and historical evidence.

Heritage Building Information Modelling

Agustan et al. (2025) examine Heritage BIM (HBIM) as a dynamic database integrating spatial and descriptive data from historical sites in Indonesia. Unlike static 3D scans, HBIM maintains relationships between structural elements, materials, historical periods, and conservation interventions. The study demonstrates how digital technology enables both preservation and interpretation—not just recording what a site looks like, but encoding what it means and how it has changed over time.

Immersive XR for Heritage Communication

Basso et al. (2024) survey the frontier of interactive XR environments supported by AI, educational Serious Games, and digital storytelling for heritage communication. Their overview identifies a shift from passive digital archives (databases of images and documents) to active engagement platforms where users interact with heritage through immersive experiences. The integration of AI dynamics—adaptive narratives, personalized content, and intelligent virtual guides—represents a new paradigm for heritage institutions.

Digital Preservation Technology Comparison

<
TechnologyBest ForLimitationMaturity
3D photogrammetryPhysical artifacts, architectureRequires physical accessMature
HBIMHistorical buildings with complex historiesHigh expertise requiredGrowing
GANs (Li, 2025)Intangible heritage visualizationGenerated images are approximationsEmerging
Text-to-image (Hsieh et al.)Lost artifacts known only from textsAccuracy cannot be verifiedExperimental
XR environments (Basso et al.)Public engagement and educationHigh development costGrowing

What To Watch

The convergence of generative AI with 3D scanning and HBIM is creating a new category of "living digital archives" that are not just records but interactive, evolving knowledge bases. Watch for UNESCO and national heritage agencies adopting AI-augmented preservation standards, and for the emerging debate about whether AI-reconstructed heritage has the same cultural value as physically preserved originals. The integration of community participation—especially from indigenous and local communities who are the custodians of intangible heritage—will determine whether digital preservation becomes a tool of empowerment or extraction.

Explore related work through ORAA ResearchBrain.

References (4)

[1] Li, Y. (2025). Driven digital preservation and rural revitalization: a multimodal framework integrating GANs and NLP for intangible cultural heritage. SPIE Proceedings.
[2] Hsieh, K., Tsaur, T.-S., & Chao, M. (2024). Cultural Heritage Meets AI: Advanced Text-to-Image Models for Digital Reconstruction and Preservation. IEEE ICCR 2024.
[3] Agustan, A. P., & Kausar, D. (2025). Digital Technology for Heritage Preservation and Interpretation in Cultural Tourism. IEEE ICCIT 2025.
[4] Basso, A., Palestini, C., & Perticarini, M. (2024). New Frontiers of Technology: Leveraging Advanced Digital Tools for Effective Cultural Heritage Communication Engagement and Preservation. ACM Proceedings.

Explore this topic deeper

Search 290M+ papers, detect research gaps, and find what hasn't been studied yet.

Click to remove unwanted keywords

Search 7 keywords →