Targeting Horror via Level and Soundscape Generation

Horror games form a peculiar niche within game design paradigms, as they entertain by eliciting negative emotions such as fear and unease to their audience during play. This genre often follows a specific progression of tension culminating at a metaphorical peak, which is defined by the designer. A player’s tension is elicited by several facets of the game, including its mechanics, its sounds, and the placement of enemies in its levels. This paper investigates how designers can control and guide the automated generation of levels and their soundscapes by authoring the intended tension of a player traversing them. Procedural content generation (PCG) is an extensive area of game research, and is often used as an effective method to reduce content creation costs and increase game longevity (Togelius et al. 2011). However, research in procedural audio is uncommon in the field (Collins 2013), likely due to the additional demands that sound often requires. Digital games are multi-faceted creative domains (Liapis, Yannakakis, and Togelius 2014), where facets such as audio, visuals, levels and game mechanics work in conjunction to create interactive digital experiences (Lopes and Yannakakis 2014). This paper investigates the interplay between level design and sound, where designers define the player experience while the system generates levels and their respective soundscapes to accommodate the designer’s intentions. The survival horror genre is unique in its heavy reliance on sound to convey negative affective states such as shock, disgust, ecstasy, fear and relief (Ekman and Lankoski 2009). It also focuses on exploration and hiding as players have limited combat ability (e.g. no weapons or limited ammunition). These complex characteristics of player affect raise important challenges for the generation of levels and soundscapes, where the focus is in evoking these types of emotions. For instance, an interesting question is how a level generator can anticipate and influence the affective state of a player, while consistently balancing feelings such as stress and relief; or how players navigate through a level under the effects of stress caused by previously encountered monsters. This paper tackles some of these challenges by exploring the generation of levels and soundscapes in the survival horror genre Copyright c © 2015, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. while also operating on a simplified player model, simulating horror gameplay during level traversal. This paper presents an extension to the Sonancia prototype (Lopes, Liapis, and Yannakakis 2015), a system capable of generating multiple facets of horror games. Levels consist of rooms in a haunted mansion that are procedurally generated via genetic search, while audio snippets are distributed throughout the rooms (i.e. level sonification). Earlier work focused on generating the architecture of the level autonomously based on the distance of a singular path. This paper focuses on the generation of levels which include added gameplay elements (i.e. monsters, quest items), and on the generation of their soundscapes, where notions of tension and suspense are used to drive the level and soundscape generative process respectively. Sonancia allows designers to define the flow of relaxation and tension through the generated level. To do this, the system requires two tension curves: the desired (designerspecified) tension curve (DTC) and the actual (level-based) tension curve (LTC). The system searches, via a genetic algorithm, for room setups where the actual tension curve more closely matches the desired one, so that generated levels and their respective soundscapes follow the designer specifications as closely as possible, while still maintaining a degree of variability.