The development of emotion perception has predominantly been studied using static, unimodal stimuli featuring the faces of young adults. Most findings indicate a processing advantage for positive emotions in children (positivity bias) and a negativity bias in adults, although these results are usually task-dependent. We created a new stimulus database comprising digital avatars from four age groups, dynamically expressing happiness, happy-surprise, anger, and sadness in visual (face only) and visual-verbal (face and voice) conditions. To determine whether previously found biases would re-emerge with this new database, we tested the arousal and valence perception of positive and negative emotions in 6- and 7-year-old children and young adults. Our results revealed high correlations between children's and adults' responses but also significant differences: children rated negative expressions as more arousing compared to adults and positive emotions as more positive than adults. Additionally, visual-verbal presentations were perceived as more arousing than visual across both age groups. In terms of valence, all participants found positive emotions as more positive in the visual condition, whereas negative emotions were perceived as more negative in the visual-verbal condition. As one of the first studies to employ dynamically multimodal emotional expressions, our findings underscore the relevance of studying developmental differences in emotion perception using naturalistic stimuli.