In this work, we develop models of cerebral venous blood flows in realistic 3D geometries and run accurate numerical 3D simulations leveraging real data. These data come from Magnetic Resonance Imaging measures and provide both morphological and physiological information on the same subject. It allows to calibrate a subject specific simulation and improve its reliability. We present the complete pipeline going from data pre-processing to the integration in the simulation framework. The 3D Finite Element simulations in the main cerebral venous vessels are coupled with Windkessel reduced models, whose parameters are chosen according to the data, to take into account the neglected network. First results are discussed and compared with literature data, opening the way to obtain reliable information difficult or even impossible to obtain in vivo in a non-invasive way. All codes are in-house openly developed ones to ensure reproducibility.