According to the principles of the invention, there is provided a system, apparatus, and method for managing media in a multimedia conferencing system according to media roles. Each media stream may be explicitly labeled with a role that describes the function or purpose of the stream, such as “people” or “content.” The labels may be hierarchical, and may include layers for media type, additional media source description, and the like, e.g., “people/presenter” or “people/presenter/video.” A policy manager is provided for managing roles, such that the multimedia conference may be more effectively presented to participants. A token management system may also be provided so that control of the multimedia conference roles can be transferred during the conference.