Relevance of emotional information varies with self-involvement. The current study was undertaken to test whether subtle facial self-resemblance is sufficient to affect attentional and affective processing of complex socio-emotional pictures. Faces digitally manipulated to resemble the participants’ (final sample: N = 21) own versus unfamiliar control faces (form and color morphs) were presented in pictures of emotionally evocative social interactions, that is, threat versus sex scenes. At stimulus onset asynchronies (SOAs) of either 300 ms or 3,000 ms after picture onset, startle responses were elicited by acoustic white noise bursts (50 ms, 105 dB), and recorded at the orbicularis oculi via electromyography (EMG). The majority of participants remained unaware of the morphing manipulation, and awareness did not affect the principal results. As indexed by inhibition of startle at short lead intervals and by picture-evoked heart rate deceleration, facial self-resemblance modulates effects of motivational salience in a context-sensitive way, increasing prepulse inhibition with threat-related, but not erotic cues. Overall, the present study suggests that visual cues of general motivational significance and of self-relevance are integrated in a fast, presumably automatic manner.