Learned vocalizations are a crucial acoustic biosignal conveying individual traits in many species. Songbirds learn song patterns by listening to a tutor song and performing vocal practice during a sensitive developmental period. However, when and how individual differences in song patterns develop remain unknown. Here, we report that individual differences in vocal output exist even at the earliest song development stage, called subsong. Experiments involving the manipulation of both breeding pairs and song tutoring conditions revealed that the parental pair combination contributes to generating familial differences in syllable duration and variability in the subsong of offspring. Furthermore, after deafening, juveniles immediately changed their subsong by shortening the syllable durations but maintained the individual variability of their subsong temporal patterns, suggesting both auditory-sensitive modification and independent intrinsic regulation of vocal output. These results indicate that the temporal patterns of subsong are not merely disordered vocalization but are regulated by familial bias with sensitivity to auditory feedback, thus generating individual variability at the initiation of vocal development.