In a mere span of two years, techniques aimed at differentiating between human-crafted and AI-generated images are swiftly losing their effectiveness. Earlier, image generation models frequently stumbled when creating images filled with text-heavy components, such as menus. For instance, menus produced by early iterations for Mexican eateries often contained misspelled dish names like 'enchuita' or 'churiros,' rendering them clearly unreliable.
