StoryVerse: Towards Co-authoring Dynamic Plot with LLM-based Character Simulation via Narrative Planning

StoryVerse establishes a novel framework for interactive narrative design that reconciles the tension between rigid authorial intent and the unpredictable nature of emergent, LLM-driven character behavior. By utilizing a structure defined as abstract acts, the system functions as a mediator that translates high-level narrative goals into concrete, context-aware character actions. This methodology effectively bridges the gap between traditional symbolic narrative planning, which often requires exhaustive hand-crafted knowledge bases, and the fluid, generative capabilities of large language models.

The core mechanism relies on an iterative planning process that continuously evaluates world states and character motivations to instantiate plot points. By employing placeholders and dynamic action sequences, the system ensures that narrative progression remains coherent even when subjected to player interventions or autonomous character decisions. This asynchronous co-creation model allows for a living story environment where the plot adapts organically to the state of the game world, providing a more responsive and immersive experience than static, pre-scripted narratives.

While the current scope focuses on the architectural integration of narrative planning and character simulation, the framework is positioned to transform how developers approach complex storytelling in interactive media. By reducing the technical overhead associated with manual narrative design, this approach facilitates the creation of sophisticated, adaptive plots that maintain structural integrity. Future advancements will prioritize the establishment of quantitative performance metrics and comprehensive user studies to validate the system’s efficacy in delivering high-quality, player-centric narrative experiences across diverse game genres.