Speech not only transmits semantic information through words and syntax, but also provides cues to a talker’s identity. Differences in a listener’s ability to recognize voices can be attributed to their language background, and in rare cases voice recognition can be selectively damaged in neurological patients. In this study we investigated a group of Korean- English bilinguals and non-Korean speakers’ ability to learn to recognize unfamiliar Korean and English talkers by voice, and to generalize to utterances not heard during training. We observed an interaction between language background and stimulus language for speed of learning, however generalization performance indicated no such interaction when compared to baseline performance. Bilinguals’ performance recognizing English (but not Korean) voices, was predicted by the age they learned English. We also observed that individuals who actively participated in music production exhibited significantly faster task learning than those who did not produce music. This study indicates that language background has a gradient effect on voice learning among bilinguals, and that non-linguistic auditory processing differences, such as music perception, impact voice identification.