This project associates multi-camera 3D modeling, physical simulation, and tracked head-mounted displays for a strong full-body immersion and presence in virtual worlds. Three-dimensional modeling is based on the EPHV algorithm, which provides an exact geometrical surface according to input data. The geometry enables computation of full-body collisions with virtual objects animated by a physical simulation. Since the algorithm is exact, it allows for consistent texture mapping and yields qualitative models. This full-body representation can thus be rendered on a distant site for telepresence. It can also be rendered into a HMD. Users see their 3D models superposed with their real bodies occluded by the HMD. Since the displays are hold in front of the eyes, the image projection is not impaired by elements of the real world. With a fixed screen, even in an immersive Cave-like configuration, users would not be able to see a virtual object in their palms, because their hands would occlude the light emitted by the displays. With our approach, users see the 3D model of their hands and the virtual object correctly positioned in their palms. It enables a first-person viewing and occlusion-free, co-located interactions. The 3D modeling system makes no assumption about the scene observed. One or several persons can stand in the acquisition space. The number of persons only affects the model quality and the computation time. We presented a similar concept, the Grimage platform, in SIGGRAPH 2007 Emerging Technologies. But the acquisition space was limited to a small volume for modeling the user's hands, and images were rendered on a fixed screen positioned behind the acquisition space, providing little immersion and only third-person interactions.