Publication Date

Spring 2024

Degree Type

Master's Project

Degree Name

Master of Science in Computer Science (MSCS)

Department

Computer Science

First Advisor

Chris Pollett

Second Advisor

Genya Ishigaki

Third Advisor

Kevin Smith

Keywords

ChatGPT, A-Frame, Three.js, Virtual Reality, 3D modeling, Code Generation, Audio Commands

Abstract

ChatGPT is a large language model that is capable of creating conversational text and functional code that can be integrated into various technologies, including computer graphics software. Currently, 3D modeling applications can be relatively difficult for novices to learn and understand due to the overwhelming amount of graphical user interfaces. However, we can remedy this issue by leveraging ChatGPT’s conversational language capabilities. Our project described in this report integrates ChatGPT with A-Frame, an online framework for developing virtual reality experiences, to create an immersive and user-friendly 3D modeling environment where users can create and modify 3D models through natural language prompts. We performed extensive testing across different temperature values that determine how repeatable ChatGPT’s outputs are. The most reliable outcomes were achieved with temperatures of 0.5 and 1.0. We discovered that the quality of the outputs usually depended on the complexity of the given prompts. This meant that simpler commands produced more precise and relevant results in comparison to complex commands.

Available for download on Friday, May 16, 2025

Share

COinS