Estimation of facial shapes plays a central role for face transfer and animation. Accurate 3D face reconstruction, however, often deploys iterative and costly methods preventing real-time applications. In this work we design a compact and fast CNN model enabling real-time face reconstruction on mobile devices. For this purpose, we first study more traditional but slow morphable face models and use them to automatically annotate a large set of images for CNN training. We then investigate a class of efficient MobileNet CNNs and adapt such models for the task of shape regression. Our evaluation on three datasets demonstrates significant improvements in the speed and the size of our model while maintaining state-of-the-art reconstruction accuracy.