Super Resolution (SR) performance has significantly increased in recent years due to the introduction of Deep Learning. In the ideal scenario, where the degradation model of a test low-resolution image does, in fact, conform with the assumed model (e.g., bicubic down-scaling) without unknown parameters like sensor noise, non-ideal point spread function (PSF), etc., supervised deep learning SR methods perform exceptionally well because they frequently rely on training data created by an assumed or predetermined degradation model. However, real low-resolution photos hardly ever fit into this perfect environment. In this study, we present a hybrid approach to the SR challenge that takes into account both external knowledge gathered by a pretrained SR network and internal information specific to a given image. This results in a hybrid network for SR that makes use of both information channels and adjusts to the provided image (and maybe its particular degradation). Experiments show that the proposed strategy works better than the state-of-the-art techniques, particularly in situations where the degradation model is unclear or suboptimal.