Kinship recognition is known in various species and evolutionarily considered a function of inbreeding avoidance and nepotism. In humans, kinship may be a crucial element of social interactions. While there are a number of studies on kinship recognition based on faces, little is known about kinship recognition in voices. Using vowel and sentence utterances, we investigated whether voices of female pairs with different degrees of relatedness (monozygotic twins, dizygotic twins and non-twin siblings) are distinguishable from unrelated female pairs (with control pairs matched to siblings for mean intra-pair age difference). Listeners’ performance to detect sibling pairs was significantly above chance (d’ = .467) and particularly high for monozygotic twins (d’ = .645). Furthermore, acoustic analyses (including F0-F4, formant dispersion, HNR, jitter and shimmer) suggested that a smaller fundamental frequency (F0) difference within a pair of voices was correlated with higher kinship judgement, in particular with simple vowel stimuli (r = -.356). Moreover, smaller differences in F1-F2 vowel space distance were associated with higher kinship judgement in sentences. These findings establish empirical evidence for human kinship recognition from voices, and provide initial evidence regarding the acoustic cues that mediate human kinship judgements from voices.