The rupture dynamics of earthquakes span a considerable range, from slow to super-shear ruptures. Rupture speed has a major influence on the generation of strong ground motions, yet our understanding of the mechanisms causing these variations in the rupture speed are still elusive. Laboratory friction experiments also bring evidences of the variability in rupture speed [Rubinstein et al., 2006; Ben-David et al., 2010]. Compared to real earthquakes, experimental ruptures occur in a controlled environment and can thus provide dense information to study the dynamics of slip events. We use a finite element method, which allows us to access detailed information on the dynamics of friction, to simulate the propagation of slip fronts at frictional interfaces. The studied setup is similar to the experimental setup used by Ben-David et al. [2010]. It consists of a block of viscoelastic material in contact with a rigid body. A velocity-weakening friction law controls the friction at the interface. We apply a shear load to the block, either on the top surface of the block or on one side. In both cases, the resulting shear tractions at the interface are non-uniform. The stress distribution presents a high concentration close to the edge when the load is applied on the side. The speed of the applied shear load, which is displacement controlled, is several orders of magnitude slower than the elastic shear wave speed. Applying a non-uniform shear loading, we observe a sequence of slip precursors, which initiate at shear levels well bellow the global static friction threshold. These precursors stop before propagating over the entire interface, and their lengths increase with increasing shear force (Figure 1). Our results are consistent with previous experimental observations [Rubinstein et al., 2007]. We also show that the velocity of the slip front varies with changing static stress state along the interface, as recently observed experimentally [Ben-David et al., 2010]. However, a simple static stress criterion is not enough to describe the propagation speed. We thus show that an energetic criterion, relating the slip front speed with the relative rise of the energy density at the slip tip is more appropriate [Kammer et al., 2012].