Artificial game narration presents unique challenges, particularly in balancing coherence with creative engagement. While recent advances have explored storytelling systems, most rely solely on textual or audio modalities, often resulting in narratives that lack contextual grounding and expressiveness. This is especially evident in non-embodied systems that struggle to maintain user engagement during gameplay narration. Despite efforts to improve narrative quality, there remains a gap in integrating embodied agents capable of delivering coherent yet expressive narrations grounded in gameplay logic. Here we present a hybrid approach that combines rule-based narration with creative storytelling elements through Nova, an embodied system that transforms structured gameplay data into spoken narratives. Nova ensures consistency with game rules while aiming to enhance engagement through expressive delivery by a physical robot. We evaluated this approach using the Artificial Social Agent Questionnaire (ASAQ), measuring Performance, Agent’s Sociability, Agent’s Personality Presence, and User Emotion Presence. Two narration styles were tested: a baseline technical descriptor and Nova’s expressive version. While no statistically significant differences were found between the two narration styles, Nova consistently received positive evaluations and showed a tendency toward higher scores in perceived performance and emotional engagement. These findings highlight the feasibility of combining structured narration with expressive delivery in an embodied format and point to promising directions for future research in artificial storytelling and human-agent interaction.