Research on Interactive Story Reconstruction Based on AI Role Dialogue and Visual Clue Analysis

This study proposes an AI role-playing and reasoning-assisted model for interactive narrative games, exploring innovative paths for interactive narratives through characterized AI. It combines the player with an intelligent driver to piece together a complete story based on the character’s dialog and visual cue analysis. The system takes memory reconstruction as the core narrative vein, allowing the player to take on the role of a mother suffering from situational amnesia. Each image uploaded by the player is interpreted as a cue from the environment, which triggers an emotionally perceptive response from an AI agent playing the role of the daughter. There is also the option to gain clues by talking to the AI-played daughter character while reasoning. To enhance immersion, the AI is designed to partially hide or distort facts during the dialog with the player in four narrative evolution stages, dynamically regulating the information credibility to simulate the uncertainty and emotional tension during human memory reconstruction. The system combines CLIP-based image similarity analysis with ChatGLMbased cue-driven dialog generation. Preliminary experiments show that players perceive the AI as emotionally appealing and that the layered narratives effectively lead them to reconstruct hidden narratives. This study demonstrates a lightweight and scalable framework for AI-assisted emotional narratives and suggests future directions including contextual memory and multimodal interaction.