The advantages of face processing over processing items in other object categories have been found in conditions both with and without visual awareness. We examines the possible mechanisms underlying the advantages of processing visible and invisible faces. Specifically, contributions from face-like configurations and face-specific local attributes are evaluated by comparing perceptual processes of three image types: genuine faces, face-like objects and non-face-like objects. Illusory face perception induced from face-like objects largely depends on information from a global configuration level, with a high tolerance for variations in local features. Comparable processing advantages for real faces and face-like objects are observed in invisible conditions using interocular rivalry and crowding paradigms. However, the visible face advantage seems to be exclusive to genuine faces. These results suggest that the face processing privilege is twofold: it is strongly triggered by global configuration without visual awareness, but with awareness it relies on facial local attributes.