Creating realistic populated virtual environments is a challenge that many graphics and VR researchers are currently tackling. There are many interesting problems to solve, such as rendering and animating large and varied crowds efficiently and realistically in believable surroundings, and creating plausible behaviours and sounds for the individual inhabitants and their environment. This is the challenge that we are addressing in the Metropolis project, where our aim is to create the sights and sounds of a convincing crowd of humans and traffic in a complex cityscape. Exploring the perception of virtual humans and crowds is also integral to our approach, through psychophysical experiments with human participants.