This thesis examines the boundaries of cinematic immersion within AI-driven virtual production environments. As interactive cinema increasingly incorporates autonomous Artificial Intelligence (AI) agents and high-fidelity digital humans, a distinct failure mode emerges: the Immersion Gap. This gap is produced when technical limitations, particularly response latency and facial animation misalignment, disrupt the emotional connection between a participant and a digital character. The study centers on a series of participant interactions with “Lacey Gordon,” a MetaHuman character developed in Unreal Engine 5 and powered by the Convai AI platform. Through these interactions, this research identifies specific thresholds at which participant immersion breaks down and proposes a set of practical Design Constraints to sustain what is here called “temporal realism”—the sense that a digital character occupies time in a believable, human way. Rather than claiming a sweeping transformation of cinema, this thesis makes a more specific argument: that when particular design conditions are met, participants are willing to engage socially with AI-driven digital characters, even while remaining aware of the underlying technology. The findings are grounded in direct observation of five participants and supported by the theoretical frameworks of André Bazin, Janet Murray, Vivian Sobchack, and Masahiro Mori.

Project documentation, including footage of the Lacey Gordon laboratory in use, is available at: https://vimeo.com/1212788182