Emotional sounds are processed within a large cortico-subcortical network, of which the auditory cortex, the voice area, and the amygdala are the core regions. Using 7T fMRI, we have compared the effect of emotional valence (positive, neutral, and negative) and the effect of the type of environmental sounds (human vocalizations and non-vocalizations) on neural activity within individual early stage auditory areas, the voice area, and the amygdala. A two-way ANOVA was applied to the BOLD time course within each ROI. In several early stage auditory areas, it yielded a significant main effect of vocalizations and of valence, but not a significant interaction. Significant interaction as well as significant main effects of vocalization and of valence were present in the voice area; the former was driven by a significant emotional modulation of vocalizations but not of other sounds. Within the amygdala, only the main effect of valence was significant. Post-hoc correlation analysis highlighted coupling between the voice area and early stage auditory areas during the presentation of any vocalizations, and between the voice area and the right amygdala during positive vocalizations. Thus, the voice area is selectively devoted to the encoding of the emotional valence of vocalizations; it shares with several early stage auditory areas encoding characteristics for vocalizations and with the amygdala for the emotional modulation of vocalizations. These results are indicative of a dual pathway, whereby the emotional modulation of vocalizations within the voice area integrates the input from the lateral early stage auditory areas and from the amygdala.