Many of the million animals dedicated to the deities in Ancient Egypt were canids. In contrast to the rare textual sources, the abundance of skeletal remains offers the opportunity to address the question of rather wild or domestic canids were mummified. However, species identification from osteological material remains problematic because it relies on a simple qualitative appreciation or traditional biometric analyses with a low discriminatory power, paired with incomplete comparative reference samples. Here we propose a new method of identification based on cranial form using a 3D landmark based geometric morphometric (GMM) approach. We built predictive methods using a large reference sample of numeric models of crania of modern canids, including a variety of domestic breeds (N=69, 38 different breeds) as well as feral dogs (N=31), and all species of wild canids present in Africa or the Near East and likely to have been present in Ancient Egypt (N=157). We then applied these methods to a sample of ancient canid remains (N=41). We compared the effectiveness of multivariate discriminant analyses based on 3D GMM to that using traditional linear morphometric measurements (LMM) commonly taken in the field. GMM provided much better results than LMM, cross-validation percentages reaching over 97.5% when determining the domestic/wild status, and 96.4% when determining the species among a reduced sample of wild canids (versus 88.2% and 85.2 % in LMM). With 3D GMM we detected the presence of dogs, but also African golden wolves and, for the first time, Near Eastern gray wolves among the mummies.