When generating an image using a GAN, some errors might appear which are obvious to the human observer, but not to the computer. By introducing this user feedback one can refine the generated image to look even more real.