Abstract
One of the least noticeable applications in the area of Procedural Generated Content is in generating narratives. Furthermore, when it tends to be used, it focuses on plot generation. Unfortunately, this results in a disregard for the potential emotional impact that storytelling may bring to table, since that, in many cases, it is the way a narrative is told that results in some sort of emotional resonance for an audience. Through a study of storytelling techniques, together with an analysis of existing narrative-driven PCG systems, and by exploring various manners of emotional evaluation in art, a system was built that takes base narratives as input and, through an emotional model, adds storytelling mechanisms to them, seeking to differentiate and maximize one of two opposite emotional moods: Happy and Dour. These narratives are then played in the Virtual Tutor system. A series of experiments consisting of four different narratives were made to discern if the emotional valence of these was successfully differentiated and maximized.