We present a 3D photography method that generates a texture-mapped three-dimensional model of a scene computed from multi-view calibrated two-dimensional photographs. Our approach first performs probabilistic space carving, which results in a 3D grid of voxel probabilities that describe the likelihood of a voxel existing in the model. We then employ a three-dimensional geodesic active surface to extract a most likely conformally-weighted minimal surface from the voxel probabilities. This surface is then polygonalized and texture-mapped, yielding in a 3D model easily rendered with standard graphics hardware.