Advancing Visual Heritage Access with AI: Use Cases for Art, Audiovisual, and Multimedia Collections

Abstract

AI4LAM is an international community advancing responsible and practical applications of AI across libraries, archives, and museums. Its collaborative Working Groups support institutions in developing shared methodologies, tools, and expertise for AI enhanced access to cultural heritage collections.

Recent use cases demonstrate the transformative potential of these approaches: for example Stanford University Libraries and the Bibliothèque nationale de France have implemented AI driven pictorial collection explorers that enable enriched discovery, visual similarity search, and new modes of engagement with large scale image corpora. Building on such initiatives, AI4LAM has established new 3D and Image Analysis WG, focusing on AI supported processing of 2D and 3D cultural heritage materials, including segmentation, classification, and advanced rendering workflows.

AI4LAM's Speech-to-Text WG further expands access to audio-visual heritage by improving transcription workflows, evaluating emerging ASR technologies, and supporting institutions in processing extensive AV collections. Building on this momentum, a new WG dedicated to audio-visual materials is currently also being established, focusing on AI supported analysis, enrichment, and access to complex AV holdings. These efforts strengthen the capacity of libraries to integrate cutting edge techniques, expand access to complex visual collections, and foster a shared, community driven roadmap for innovation in the cultural heritage sector.