Teaching using multimedia is a field where social robots can contribute in a great manner. Some works in humancomputer interaction and multimedia learning demonstrated that synthesized voice impairs user's learning. In this work, we investigate the importance of the system's voice (synthesized vs. human), of the embodiment (robot vs. tablet), and of the user's gender and personality on learning nutrition and healthy eating tips. The results obtained with the Kompai robot (developed by Robosoft) show that the performance on learning is better with the human voice. Moreover, the results show that user's personality plays an important role in learning. Individuals with high Neuroticism score performed better in the multimedia learning session than individuals with low Neuroticism score. Also, the stress of male participants was higher in the condition with the robot and synthesized voice than in the other conditions These findings can be used for better developing teaching systems tailored to the user's profile.