Project description

The Relic project is a prototype driven research project aimed at exploring how designers can integrate generative content in their games without adding too much developmental complexity or technical overhead. The project consists of two game prototypes: a text-based dungeon crawler and a 2D game built in Unity.

A core focus on the Relic project is to identify issues faced by smaller independent (“indie”) games studios in applying generated content in games. Indie games have a long history of creatively applying procedural content generation, and we believe that generative AI can act as a major accelerator for the capabilities of indie studios. However, adding generative AI to games projects requires that one accepts a certain trade-off between control and content generation, which may lead to certain design challenges. Furthermore, many current generative AI systems are rife with ethical issues of artistic integrity and copyright issues.

The Relic project focuses on providing design guidance to indie studio on how to apply generative content in games. Although generative AI is currently revolutionizing much of how we create and consume content, there is still comparatively little knowledge available on just how we should apply generated content.

In order to alleviate some of these issues, the Relic project takes a broad approach to generative content and uses both traditional procedural content generation techniques, as well as applying new and cutting-edge generative artificial intelligence techniques. This allows us to be flexible in how we apply different forms of generative techniques, and also allows us to combine generative and hand-crafted content. This approach also allows us to mimic how these technologies would be applicable in a small indie games studio, providing us with ecologically relevant findings around how generative techniques can be applied.

The first public version of Relic 2D is slated for release in late spring 2024.

Ongoing threads of research

The Relic project encompasses a number of independent but related threads of research.

Narrative Generation
Relic started as a project to investigate how to apply generative AI to create interesting and engaging narratives. Fully generated narratives, either procedurally or through generative AI, has long been a goal for those of us involved in narrative generation. However, as evidenced by OpenAI’s own evaluation of their GPT-4 system, even the most impressive and recent attempts are somewhat limited in what types of narratives they can produce. The Relic project takes a different approach where we focus on the generation of narratives that support the main narrative. By keeping the main narrative human-authored we are also able to better control the overall direction of the game’s story. However, using generative AI to create supporting narratives that tie into the main narrative we can add a significant amount of content to the game without adding a commensurate amount of work for the development team.

Non-Player Character Design
This thread of the Relic project is a continuation of my previous research on the social believability and design of Non-Player Characters (NPCs). We are working on expanding the NPC typology that I created with my colleagues. The aim is to better describe the behaviors and design elements associated with different types of NPCs. The findings from this study will inform the creation of generative algorithms for NPC creation.

Accessibility
The Relic 2D prototype is being used as a platform to test the impact of accessibility techniques to aid players with visual impairments. This study was performed in Spring 2024 by Jhané Dawes.

Credits

Final project members
Dr Henrik Warpefelt - principal investigator
Joshua Whorton - project lead, NPC AI programmer
Joel Metukmebong - project development lead, combat design & programming
Stephen Pangilinan - NPC art and design
Ryan Whisenhunt - programming (gameplay, systems)
Anaiya Tucker - puzzle design, user research
Angela Okafor - artist
Donovan McGregor - music
Jhané Dawes - accessibility research

Former project members
Justin Graham - project lead, lead artist
Katie Buell - generative content programmer
Cameron Cederquist - generative content programmer
Nicholas Goolsby - accessibility research
Lucas Kim - character design
Emily Espinoza - user research
Victoria Davis - game art
Margarita Marquez - user research
Gabriel Craven - generative content design