Generating multifaceted content in games : a study on levels and sound

Abstract

Audio is an often overlooked aspect of digital games as it consistently remains unnoticeable to players entwining perfectly with the on-screen experience. Good sound design is considered to enhance the interactive experience, never overshadowing either the narrative or visuals of the game. Audio can be used to convey information about the world nearby (e.g. footsteps, voices), simulate real-world soundscapes (e.g. rain or wind), and enhance the experience of the on-screen drama and gameplay sequences. Narrative-heavy games, such as Amnesia: The Dark Descent (Frictional Games, 2010), rely heavily on emotional progressions that unveil during play. Players are tasked to push through situations that evoke fear, relief or even confusion, perfectly conveyed by the synchronization of both visual and audio facets; for instance, sombre music slowly fades in as players gradually step into an unlit room. Procedural content generation is a popular field within digital game AI research, however most work tends to focus specifically on the creation of virtual spaces. This thesis argues that this type of generation can be quite limiting, especially for certain types of genres. By interweaving different facets in the content creation process, as is done in actual game development, can potentially provide a deeper and a more exciting experience for the players. This thesis introduces a system called Sonancia, an autonomous content generator capable of constructing horror levels and their respective soundscapes. A number of AI techniques have been used and tested for both the construction of levels that adapt to a user (or a machine) defined progression of emotion, and the creation of soundscapes that adapt to these emotional progressions. Sonancia is evaluated thoroughly via extensive user studies