In this paper, we present a modeling methodology to couple the cardiac conduction system to cardiac myocytes through a model of Purkinje-ventricular junctions to yield fast and realistic electrical activation of the ventricles. A patient-specific biventricular geometry is obtained from processing computed tomography scan data. A one-manifold implementation of the fast marching method based on Eikonal-type equations is used for modeling heart electrophysiology, which facilitates the multiscale 1-D-3-D coupling at very low computational costs. The method is illustrated in in-silico experiments where we analyze and compare alternative pacing strategies on the same patient-specific anatomy. We also show very good agreement between the results from the proposed approach and more detailed and comprehensive biophysical models for modeling cardiac electrophysiology. The effect of atrioventricular delay on the distribution of activation time in myocardium is studied with two experiments. Given the reasonable computational times and realistic activation sequences provided by our method, it can have an important clinical impact on the selection of optimal implantation sites of pacing leads or placement of ablation catheter's tip in the context of cardiac rhythm management therapies.