Morphing wings are widely regarded as one of the most promising technologies for improving the aerodynamic performance in large civil aircraft. The controllable adjustment of the wing trailing edge in response to external operative conditions naturally improves the aerodynamic efficiency. However, the deformation of trailing edge plays a crucial role in flow separation, leading to significant changes in lift, drag, and stability. Therefore, quickly and accurately predicting the unsteady separated flow around the airfoil is essential for optimizing aerodynamic performance and ensuring stability under different flight conditions. This paper presents a hybrid model that improves the prediction of unsteady separated flows over an airfoil using reduced-order modeling (ROM) and super-resolution techniques. Delayed detached eddy (DES) simulations of the OAT15A airfoils with various trailing edge deformations are conducted and used as the dataset for the model. The model decomposes the temporal flow field data to obtain low-order reconstructed flow fields and their temporal parameters. The Vision Transformer enhances the spatial details of the low-order reconstructed flow field with physical constraints, while the reduced-order model is used for time stepping. By combining reduced-order modeling and super-resolution techniques, the model improves flow prediction resolution, capturing fine-scale features often missed by conventional models. This enhancement provides a more accurate representation of the flow, offering valuable insights for airfoil design and optimization.