It is commonly observed that acoustic echoes hurt perfor-mance of sound source localization (SSL) methods. We in-troduce the concept of microphone array augmentation withechoes (MIRAGE) and show how estimation of early-echocharacteristics can in fact benefit SSL. We propose a learning-based scheme for echo estimation combined with a physics-based scheme for echo aggregation. In a simple scenario in-volving 2 microphones close to a reflective surface and onesource, we show using simulated data that the proposed ap-proach performs similarly to a correlation-based method inazimuth estimation while retrieving elevation as well from 2microphones only, an impossible task in anechoic settings.