Previous work has found faster and more accurate recognition of faces learnt from moving sequences than from static images. However, it is unclear whether the motion advantage can be generalised to other-race face learning. In two experiments, we examined the motion advantage for same- and other-race faces in the context of a delayed visual search task and investigated the importance of face repetitions and clip length at learning. Participants learnt faces as a static image or a moving clip and then searched for these target faces in visual arrays. Same-race target faces were located faster and more accurately than other-race targets. A dynamic search advantage was revealed for both same- and other-race faces, whereby search latencies were shorter, and accuracy higher for faces learnt in motion. Furthermore, differing clip lengths and face repetitions during familiarisation yielded the same dynamic advantage (Experiment 2), suggesting that motion provides a robust and valuable cue for the identification of both same- and other-race faces.