A Text-to-Game Engine for UGC-Based Role-Playing Games

Lei Zhang,Xuezheng Peng,Shuyi Yang,Feiyang Wang

2024-07-11

Abstract:The shift from professionally generated content (PGC) to user-generated content (UGC) has revolutionized various media formats, from text to video. With the rapid advancements in generative AI, a similar shift is set to transform the game industry, particularly in the realm of role-playing games (RPGs). This paper introduces a new framework for a text-to-game engine that utilizes foundation models to convert simple textual inputs into complex, interactive RPG experiences. The engine dynamically renders the game story in a multi-modal format and adjusts the game character, environment, and mechanics in real-time in response to player actions. Using this framework, we developed the "Zagii" game engine, which has successfully supported hundreds of RPG games across a diverse range of genres and facilitated tens of thousands of online user gameplay instances. This validates the effectiveness of our frame-work. Our work showcases the potential for a more open and democratized gaming paradigm, highlighting the transformative impact of generative AI on the game life cycle.

Artificial Intelligence,Computation and Language,Multiagent Systems

What problem does this paper attempt to address?

The paper attempts to address several challenges present in the traditional role-playing game (RPG) development process, including:

1. **High development costs**: Traditional RPG development requires multidisciplinary professional skills such as scriptwriting, character design, game mechanics, programming, and graphic design, making the development process both expensive and time-consuming.
2. **Poor flexibility**: The traditional RPG development process is relatively rigid, and changes in one area often trigger a chain reaction in other areas, leading to extended development cycles.
3. **Low player freedom**: Traditional RPGs typically offer linear or branching paths and lack the ability to dynamically adjust game content based on player choices, limiting player freedom and control.
4. **Limited content expansion**: Traditional RPGs rely on fixed engines, making it difficult to dynamically generate new content based on player interactions, resulting in a limited and repetitive gaming experience.

To address these issues, the paper proposes a new framework based on a text-to-game engine that utilizes foundational models to convert simple text inputs into complex interactive RPG experiences. This engine can dynamically render game stories and adjust game characters, environments, and mechanics in real-time in a multimodal format in response to player actions. Through this framework, the authors developed the "Zagii" game engine, which successfully supported hundreds of different types of RPG games and facilitated tens of thousands of online user game instances, validating the framework's effectiveness. The paper demonstrates the transformative impact of generative AI on the game lifecycle, promoting a more open and democratized gaming paradigm.