Integrating LLMs and Extended Reality for Interactive History-Telling Performances

This paper presents a novel approach to cultural heritage enhancement that integrates generative artificial intelligence, extended reality, and theatrical performance. Working at the intersection of computer engineering, history, and dramaturgy, we report on an experience carried out at the Carlo V Castle in Lecce, Italy, where a historically accurate virtual representation of the 16th-century architect Gian Giacomo dell'Acaya confronted human actors and live audiences. The technical architecture, which supports natural language generation with period-appropriate language models, noninvasive projection techniques for historic sites, and multimodal interaction capabilities, is described in detail. Evaluation through public questionnaires (n=90, 25% response rate) and stakeholder interviews revealed a high degree of satisfaction (4.2/5) and identified both strengths and challenges of the approach. The results demonstrate the potential of AI-enhanced performances to make historical narratives more accessible and engaging, particularly for younger audiences, while preserving the essential human elements of theatrical art. This work contributes to the emerging field of computationally enhanced cultural performance by proposing a methodological framework that respects historical authenticity while embracing technological innovation.