Papers

A designed world, however beautiful, is still a script. Open-world games have mastered immersion through narrative, art, and music, yet their inhabitants remain bounded by authored conversations and relationship branches. Players want more: randomness, conversation without rails, relationships that move. Large language models have begun to loosen the script, but between a model that talks and a character that lives, the distance is still wide. We present the Non-Player Character Bank (NPCBank), an ecosystem for crossing it. NPCBank Data grounds characters in source evidence and turns them into structured dialogue cases; NPCBank Web makes the resulting characters, evidence, and evaluation contracts open to inspection and annotation; and NPCBank Lab presents a demo game for testing interactions with model-based NPCs. Beneath all three runs the craft: twenty-one dialogue capabilities, an ontology and schema for contract-based dialogue cases, and an agent that builds characters and writes their cases from source material. Building on NPCBank, we formulate three challenges: whether models can act on supplied state while preserving character and knowledge boundaries; whether capable models can run effectively on consumer GPUs; and whether they can sustain role-play across extended interactions as context, relationships, and conversational pressure evolve. We release NPCBank as a shared foundation for building and studying interactive characters across game AI, conversational agents, and computational character modeling.