Vocal communication is of major social importance in pigs. Their auditory sensitivity goes beyond the intraspecific level; studies have shown that domestic pigs are sensitive to and can learn to recognise human voices. The question of which prosodic features (intonation, accentuation, rhythm) of human speech may matter to this recognition, however, remains open. A total of 42 piglets were allocated to three experimental groups. Each piglet was submitted to three choice tests, during which different pairs of sounds were broadcast. Each group was first offered a choice between an unmodified (neutral) human voice and a background noise, in order to verify the attractiveness of human voice. We found that piglets could distinguish human voice; they gazed more rapidly (P < 0.05) and for longer (P < 0.05) in the direction of the human voice than in the direction of the background noise. Group 1 was then submitted to artificially modified voices: low vs high-pitched, and then slow vs rapid rhythm. Group 2 was submitted to artificially modified voices with a combination of these features: rapid and high-pitched vs slow and low-pitched, and then slow and high-pitched vs rapid and low-pitched. Group 3 was submitted to naturally recorded voices coding for different emotions (happiness vs anger) and then different intonations (interrogation vs command). We found that piglets approached the loudspeaker broadcasting the rapid rhythm (6 s (2–32)) more rapidly than the loudspeaker broadcasting the slow rhythm (33 s (15–70); p < 0.05). They also spent more time near the loudspeaker broadcasting the “high-pitched and slow” voice (86 s (52–110)) than near the one broadcasting the “low-pitched and rapid” voice (29 s (9–73); W = 86, P < 0.05). In sum, the sensitivity of piglets for human prosody was moderate but not inexistent. Our results suggest that piglets base their responses to human voice on a combination of prosodic features.