Presentation Information
[2Yin-A-31]Real-time BGM generation system based on image recognition of party composition and map features in RPG
〇Takuma Sugano1, Yoshiharu Maeno1 (1. Meiji University)
Keywords:
BGM,Generation AI,Game
This study proposes a system that extracts party composition and map features in real-time from role-playing game (RPG) screens using image recognition, and automatically generates the background music (BGM) for that situation. This system uses image analysis to extract party composition and map features, then employs generative AI to construct prompts for music generation. Inputting these prompts into music generation AI produces BGM matching the party's state, aiming to enhance immersive feeling. We confirmed the possibility that the system-generated BGM is perceived as natural and unobtrusive by human listeners.