Content preparation systems and methods are disclosed that generate scenes used by an interactive role performance system for inserting a user image as a character in the scene. Original media content from a variety of sources, such as movies, television, and commercials, can provide participants with a wide variety of scenes and roles. In some examples, the content preparation system removes an original character from the selected media content and recreates the background to enable an image of a user to be inserted therein. By recreating the background after removing the character, the user is given greater freedom to perform, as the image of the user can perform anywhere within the scene. Moreover, systems and methods can generate and store metadata associated with the modified media content that facilitates the combining of the modified media content and the user image to replace the removed character image.