Contemporary visual environments bombard us with hundreds of face images every day, and this places a non-trivial demand on long-term memory. However, little is known about what makes certain faces remain in our memories, while others are quickly forgotten. To establish a basis for face memorability exploration, we assembled a database of 8,690 face photographs from online sources, spanning diverse face and image characteristics. Workers on Amazon's Mechanical Turk were asked to identify repetitions within a stream of these stimuli. Variations in image memorability (hit rates, false alarm rates, and their interactions) were reliable across participants, suggesting that face images may have different intrinsic levels of memorability. We discuss future directions in using this database to quantify face photograph memorability, as well as potential scientific and commercial applications.