Affective states induced through robotic speech play a crucial role in communication between robots and humans. The embodiment hypothesis suggests that the perception of and behavior toward robots is heightened by their physical presence, yet little is known about how this embodiment influences affective responses in humans. Our laboratory study involved a depicted and an embodied robot to examine whether a robot's affective speech elicits corresponding arousal and emotional valence in humans. The findings revealed three key insights: First, positive and negative speech significantly increased arousal, with positive speech also resulting in a more positive emotional valence. Second, participants experienced higher arousal levels when interacting with an embodied robot compared to a depicted one. Third, there was a clear preference for future interactions with embodied robots over their depicted counterparts. These results emphasize the importance of both speech and embodiment in shaping human emotional responses in human-robot interactions.