Multimedia documents such as photographs, videos, texts, and web pages are essential for understanding how urban environments evolve, but exploring them in relation to a 3D city model remains difficult because spatial, temporal, and thematic relations must be handled simultaneously. This paper investigates how a design space for such visualizations can be constructed without reducing either the urban context or the informational richness of the documents. We adopt a design-driven methodology based on iterative prototyping: visualization propositions are developed from initial requirements, tested through simulation, confronted with user needs, and refined as new needs and limitations emerge. The paper first structures the corpus in a spatio-temporal-thematic model, then formalizes the main culling and display functions required for exploration. On that basis, we derive a family of complementary visualization propositions for tasks including spatial selection, temporal filtering, georeferenced media display, non-georeferencable document organization, link visualization, and document quantity analysis. Rather than proposing a single interface, the contribution is a structured design space and a set of concrete visualization propositions for multimedia exploration in 3D urban environments.