Predator presentation experiments are widely used to investigate animal alarm vocalizations. They usually involve presentations of predator models or playbacks of predator calls, but it remains unclear whether the two paradigms provide similar results, a major limitation when investigating animal syntactic and semantic capacities. Here, we investigate whether visual and acoustic predator cues elicit different vocal reactions in black-fronted titi monkeys (Callicebus nigrifrons). We exposed six groups of wild titi monkeys to visual models or playbacks of vocalizations of raptor or felid. We characterized each group’s vocal reactions using sequence parameters known to reliably encode predatory events in this species. We found that titi monkeys’ vocal reactions varied with the predator species but also with the experimental paradigm: while vocal reactions to raptor vocalizations and models were similar, felid vocalizations elicited heterogeneous, different reactions from that given to felid models. We argue that subjects are not familiar with felid vocalizations, because of a lack of learning opportunities due to the silent behaviour of felids. We discuss the implication of these findings for the semantic capacities of titi monkeys. We finally recommend that playbacks of predator vocalizations should not be used in isolation but in combination with visual model presentations, to allow fine-grained analyses of the communication system of prey species.