A story-based video game contains many characters. The majority are virtual characters controlled by artificial intelligence. In recent years, virtual character artificial intelligence has developed slower than other aspects of video games, such as graphics, mainly due to the cost of scripting complex and believable virtual characters. To tackle this bottleneck in content creation, this research proposes a new Tiered Behavior Architecture model for controlling the behaviors of virtual characters. For local scenes, techniques such as Behavior Capture with Hidden Markov Models, which has been evaluated by user studies that validated its success in generating fine-grained behaviors, can be used to fulfill the roles. At a larger scale, a hierarchical cyclic scheduler determines the general circumstances, schedules, and objectives of the virtual characters as well as the roles that will accomplish these objectives. This paper describes experiments and user studies that validate this model.