Active vision exploits the ability of robots to interact with their environment, towards increasing the quantity / quality of information obtained through their sensors and, therefore, improving their performance in perception tasks. Active face recognition is largely understudied in recent literature. In this paper, we propose an active approach that utilizes facial views produced by facial image rendering. The robot that performs face recognition selects the best candidate rotation around the person of interest by simulating the results of such movements through view synthesis. This is achieved by passing to the robot's face recognizer a real world facial image acquired in the current position, generating synthesized views that differ by $\pm\theta^{\circ}$ from the current view. Then, it decides, on the basis of the confidence of the recognizer, whether to stay in place or move to the position that corresponds to one of the two synthesized views, so as to to acquire a new real image. Experimental results in two datasets verify the superior performance of the proposed method compared to the respective static approach and an approach based on the same face recognizer that involves face frontalization with synthesized views.