In this paper, we propose a two-step GAN approach for retinal image generation, aimed at addressing the challenge of limited availability of medical images. Our method involves first generating vessel trees, which are then translated into realistic retinal images using the Pix2PixHD architecture. This two-step process leverages the strengths of Generative Adversarial Networks (GANs) to produce high-quality synthetic retinal images that can be used for medical research, training, and diagnostics. By utilizing this approach, we demonstrate the potential of GANs to enhance the availability of medical images, thereby supporting advancements in medical imaging and machine learning applications in healthcare. The usefulness of this method was demonstrated by training a segmentation network using the generated images. Therefore, we trained a U-Net network using the synthesized data and compared it with real data.