SMU Guildhall Graduate Catalog 2025 | Page 28

Using Generative AI to Procedurally Populate Assets in an ASCII Level

Tianxing JIANG

Level Design

Using Generative AI to Procedurally Populate Assets in an ASCII Level

For my thesis, I explored how Generative AI can be integrated into the game development process to procedurally populate decorative assets in a level. My goal was to create a pipeline that could enhance the efficiency of the aesthetic milestone( where designers typically finalize environmental visuals) without compromising the quality of player experience. The project centered on Left 4 Dead 2— where I designed a full campaign level, implemented AI-assisted decoration, and then analyzed the results through structured playtests and surveys.
I chose this project because I wanted to merge my interests in level design and AI tools in a way that could benefit the broader game industry. It took approximately three months to complete, including preparation, implementation,
and playtesting. During this time, I significantly improved my skills in prompt engineering, level editing, and data analysis, while deepening my understanding of how iterative design and AI interaction can streamline production workflows.
This project could be highly valuable to the game industry, especially for small or indie teams with limited access to artists. It demonstrates a scalable and flexible way to enhance level decoration without draining resources, allowing designers to maintain creative control while accelerating production with AI assistance.
28 LEVEL DESIGN