We can use the characteristic way a person moves their face and head ("dynamic facial signatures") as a cue to identity. Theoretically, we should have pre-existing representations of the way a familiar face moves, making it easier to match the movement of familiar than unfamiliar faces. However, few studies have directly compared the benefits of movement for familiar and unfamiliar faces. It is also unclear whether the use of dynamic facial signatures depends on the type of movement, or a particular face area. In this study, we investigated the movement advantage for famous and unfamiliar faces using a sorting task. Participants sorted groups of moving or static shape-normalized point-light-displays (PLDs), using either rigid head movement (e.g. nodding, tilting), non-rigid face movement (e.g. smiling, talking) or combined rigid and non-rigid movement. In Experiment 1, standard PLDs were used. In Experiment 2, the PLDs included eyes, while in Experiment 3, they included the teeth and tongue. Accuracy scores were divided by the average number of times clips were viewed. Famous and unfamiliar faces were sorted equally well overall. Famous faces showed a movement advantage for combined and non-rigid clips, but not rigid clips. The results suggest that participants were using mouth information: famous face PLDs with mouths were sorted better than standard PLDs or PLDs with eyes. Like famous faces, unfamiliar faces also showed a movement advantage for combined motion. Unlike famous faces, unfamiliar faces were sorted equally well from standard PLDs and those with mouths or eyes. Overall, these results show that both famous and unfamiliar faces can be sorted based on dynamic facial signatures. Sorting famous faces may rely more on non-rigid movements of the mouth region than on the eyes or rigid motion, whereas sorting unfamiliar faces is best when both rigid and non-rigid movement are present.