cultuur.ai Belgium

KMSKA describes an internally developed prototype that lets visitors to Studio Rubens ask spoken questions about practical matters and art history. Whisper converts speech to text, a language model loaded through Ollama processes the question, and Piper TTS reads out the answer. By developing the application locally, the museum retains control of its data. The source describes a working concept still in preparation, with limitations in available computing power and Dutch-language processing; no public launch date is given.

In Ways of Seeing, the Vlaams Architectuurinstituut compares how architects, writers and AI look at three photographic archives from its own collection. Image recognition and generative AI each feature in the installation, which also questions biases and colonial perspectives in AI analyses. The exhibition runs until 31 January 2027 at DE SINGEL in Antwerp.

visit.brussels is working on a free BYOD solution for AI-assisted text and audio translation in Brussels museums. The project aims to lower language barriers with real-time translation in the 24 official EU languages and other locally spoken languages, improving accessibility, inclusion and visibility for the museum offer in the Brussels-Capital Region.

De Vlaamse Regering has given final approval to a draft decree that makes European rules on transparent political advertising, media freedom and artificial intelligence applicable and enforceable in Flanders. Of particular relevance to the media and cultural sectors, users must be able to see when programmes or content have been artificially generated or manipulated, while the Vlaamse Regulator voor de Media will oversee compliance.

meemoo shows how artificial intelligence makes audiovisual heritage easier to find and reusable. Within the Hermes project, facial recognition in photos, audio classification, voice recognition and automatically generated titles and summaries are being tested, so that hidden information in large archive collections becomes visible more quickly to heritage partners, researchers and users of hetarchief.be.

VRT looks back on three years of Seeds & Growth for Media, an innovation programme in which startups, creative entrepreneurs and Flemish media companies could test new applications more quickly. Within VRT, this led, among other things, to experiments with artificial intelligence for Radio 1, where audio from broadcasts and podcasts is converted more intelligently into transcripts, fragments and social media content for editors.

meemoo describes the ECHOLOT project, in which fifteen European partners develop open tools and work processes to more easily share, enrich and reuse cultural heritage via Wikimedia Commons and Europeana. For Flemish and Brussels collections, AI is used, among other things, for automatic summaries, descriptions and image recognition, with tests on fine arts and performing arts at partners such as KMSKA, M Leuven, Musea Brugge, Mu.ZEE, Opera Ballet Vlaanderen and De Munt.

The Vrije Universiteit Brussel is a partner in AIM-PRO, a three-year Erasmus+ project that aims to strengthen responsible AI literacy and future-oriented digital skills in European education. The VUB brings expertise in educational technology and pedagogical innovation and helps develop a human-centred, multidisciplinary approach to AI education.

BRUSK presents 'Refik Anadol. Latent City' a groundbreaking exhibition by the internationally renowned new media artist Refik Anadol. The installation was created with artificial intelligence and uses historical and real-time data from Bruges as the basis for an immersive total experience.

Charter on social media use of the Brussels GCs

Artist, programmer and engineer Kasper Jordaens’s Lost In Diffusion project seeks to better understand the role of algorithms and AI models in relation to a museum collection. AI has proved valuable in turning heritage data into accessible, interconnected and understandable information, yet its potential to generate meaningful new data remains largely untapped. Using a collection database containing metadata and images of the museum’s design objects, Jordaens retrained a diffusion model¹. The aim was both to help the machine understand the nuances of design and to create a model that could act as a co-agent alongside a designer, rethinking the relationship between people and machines in the design process.