We’re building a web-based murder mystery game where players don’t choose dialogue options—they interrogate suspects directly using natural language. Each character has a hidden profile (motives, knowledge, contradictions), and the system tries to maintain consistency while allowing open-ended questioning. Players naturally start testing inconsistencies, building hypotheses, and treating conversations like investigation rather than chat.

We built this as an experiment to explore what happens when you remove traditional dialogue trees from narrative games entirely. Instead of selecting options, players interrogate suspects directly in natural language. Each character has hidden knowledge, motives, and contradictions, and the system tries to keep responses consistent while still allowing open-ended questioning. What surprised us most is that players don’t treat it like a chatbot — they start behaving like real investigators, testing inconsistencies and forming hypotheses. We’re still early and actively tuning the balance between freedom and structure. Would love feedback on one key question: Does “free interrogation” feel more immersive than traditional branching dialogue, or does it need more structure to stay meaningful? Happy to answer anything about the design or system!