ActiveCube is a novel user interface which allows intuitive interaction with computers. ActiveCube allows users to construct and interact with Three Dimensional (3D) environments using physical cubes equipped with input/output devices. Spatial, temporal and functional consistency is always maintained between the physical object and its corresponding representation in the computer. In this paper we detail the design and implementation of our system. We describe the method we used to realize flexible 3D modeling by controlling the recognition signals of each face in each cube. We also explain how we integrated additional multimodal interaction options by a number of sophisticated I/O devices and by the inclusion of a second microprocessor in our cubes. We argue that ActiveCube, with its current real-time multimodal and spatial capabilities, is ready to enable a large range of interactive entertainment applications that were impossible to realize before.