This paper examines the representation of humans in large-scale, networked virtual environments. Previous work done in this field is summarized, and existing problems with rendering, articulating, and networking numerous human figures in real time are explained. We have developed a system that integrates together some well-known solutions along with new ideas. Models with multiple level of details, body-tracking technology and animation libraries to specify joint angles, efficient group representations to describe multiple humans, and hierarchical network protocols have been successfully employed to increase the number of humans represented, system performance, and user interactivity. The resulting system immerses participants effectively and has numerous useful applications.