Self-face perception plays an important role in self-consciousness and personal identity as well as in social exchanges and well-being. Despite its significance, little is known about how individuals represent their faces internally. This study explored mechanisms of self-face perception in three experiments. First, participants chose from two images (self-face image vs. self-image with manipulated facial features) which one was their veridical image and which one they liked most. Afterwards, participants could (digitally) manipulate their facial features to increase their attractiveness (either to themselves or to an imagined other/s). Results showed that self-face recognition was better when veridical faces were paired with 'clones' with larger facial features or when all facial features were enlarged concurrently. Moreover, up to half of the participants preferred smaller noses and larger eyes and manipulated their self-images accordingly. State (but not trait) self-esteem was inversely correlated with eye, mouth and nose size manipulations made to increase one's attractiveness. The results indicate that a certain tolerance for error in self-face recognition might be required to maintain a consistent facial identity during one's lifespan. The discovered preference for neotenous features and discrepancies between one's perceived and one's veridical face and their link to state self-esteem are discussed.