To reach the diffraction limit of large aperture ground-based telescopes is a big challenge. There are several methods for image reconstruction, such as Multi Object Multi Frame Blind Deconvolution (MOMFBD) or speckle reconstruction. However, these methods are limited in its success, require a lot of computational power and are therefore very time consuming. The reconstruction of a single high-resolution image can take several minutes up to days! We provide here a deep learning approach to reconstruct an original image burst to a single high-resolution image in real time.