This paper presents a novel conversational engine for non-player characters (NPCs) in role-playing games, designed to deliver adaptive, personality-driven dialogue without the need for training and maintaining custom models. The system integrates dynamic sentiment-based tone modulation with a dual-memory architecture that combines short-term conversational context and long-term narrative history. Advanced decay and cooldown mechanisms are implemented to gradually diminish the influence of older interactions and sentiments, thereby enabling smooth transitions in NPC behavior. Our approach dynamically recalibrates NPC responses based on real-time player inputs, world state, and historical interactions, resulting in immersive and context-aware dialogue that is highly humanlike. Coupled with the Generative AI models, our framework generates NPC responses that convincingly emulate actual personalities and emotions, while maintaining minimal reliance on the underlying large language model for fine-tuning. Experimental evaluations demonstrate that the engine produces coherent, emotionally consistent interactions, significantly enhancing the realism of NPC communication in gaming environments. This work contributes a scalable and technically robust framework that bridges the gap between traditional scripted dialogues and modern adaptive storytelling.