Create Ice Cream: Real-time Creative Element Synthesis Framework Based on GPT3.0

Lei Huang, Xing Sun

Abstract: Elemental synthesis, the combination of two objects to create a new object, has several limitations in past video games, including incomplete freedom, unreasonable results, weak repeatability, and limited content. However, with the GPT3.0 natural language model (GPT), these limitations can be overcome. GPT enables meaningful feedback for all combinations, ensures logical results, adds randomness, and vastly expands the amount of game content. We present a Real-time Creative Element Synthesis Framework based on GPT (RCESG) and designed an element synthesis game based on GPT: Create Ice Cream to demonstrate the framework. The game allows players to freely combine elements to create new content, aided by GPT's generation of the synthesis result based on physical and chemical properties, conceptual association and combination principles, and other logical reasoning rules. The game also includes different modes, such as Quick Pass Mode and Story Mode to enhance the gameplay experience. The RCESG framework can effectively bring element synthesis in video games with high freedom, high logic, repeatable playability and more content.