Consumer grade VR now allows embodying full-body avatars to act within the virtual world. Realism has been shown to be an important factor for the plausibility and self attribution of an avatar. However, the role of self-representation, diverting from the actual user’s appearance, is not clear yet. In this work, we compare a generic, a self-configured, and a photo-generated full-body avatar with similar realism level in terms of embodiment and user experience. We conducted a study with thirty participants, that revealed higher ratings of attractiveness, valence, and threat reaction (as an indirect measure for embodiment) in the self-configured condition, compared to the photo- generated avatar. However, direct measures for embodiment remained inconclusive. These results motivate further research in the field of self-configured avatars, considering the feedback about avatar diversity that is still worthy of improvement.