This work represents a first attempt at documenting the benefit of training preschool children in inferring the character’s emotion from natural stories. In one condition, preschoolers were trained to make emotional inferences from televised stories (i.e., video training), whereas in the other condition, they were trained to infer the character’s emotion from auditory stories (i.e., audio book training). Trained preschoolers were compared to children that did not participate in any training sessions (i.e., no training condition). At three separate times (i.e., before the training session, immediately after, and six weeks later), the performance of trained and untrained children was assessed from both televised stories and auditory stories. Overall, the results revealed the circumstances under which preschoolers can benefit from very simple interventions. When their ability to make emotional inferences was assessed with televised stories, trained children significantly outperformed untrained children. Moreover, the gain observed for trained children was durable over time (i.e., follow-up) and this benefit was independent of the type of stories used during the training sessions since children trained solely from audio book did not differ from children trained with video only.