Content Creation Tools for Autonomous AR Characters

Abstract

Traditionally, character creation tools require their users to be skilled in 3D modelling or programming. This thesis describes the creation of a tool which allows users to create believable characters as pictured in their mind in an Augmented Reality (AR) environment without any necessary experience in 3D character creation or programming knowledge. This tool guides the user through the following steps: designing the character’s visual appearance, determining in-framework Behaviors, defining personality, and defining character interests and background. The user can define characters’ relationships to each other, and afterwards, watch characters interact and speak with each other. The tool is created in Unity for iOS, and uses GPT-3.5 Turbo to generate character dialogue, as well as Google TextToSpeech to give each character a voice. A qualitative and quantitative user study is performed to test the tool, with a median SUS score of 72.5, which indicates that users found the tool easy to use. More work needs to be done to improve the tool’s creativity support and the believability of the characters created, but this tool lays a good foundation for future work.