GPT Image 2.5 Craiyon
retained text
Prompt: Create a realistic screenshot of the Craiyon.com website showing an image generation result for the prompt “PROMPT HERE”. The screenshot must look like an authentic old Craiyon/DALL·E Mini webpage from around 2021–2022, with the recognizable Craiyon interface, prompt input field, generation button, webpage layout, and a 3×3 grid containing nine separate 256×256 AI-generated images. The nine images should look like genuine outputs from an early primitive text-to-image AI similar to DALL·E Mini: extremely low-fidelity, blurry, muddy, melted, oversmoothed, incoherent, uncanny, artifacted, distorted, and imperfect. The AI should appear to have weak semantic understanding and limited visual knowledge. Faces should be especially poorly generated, with asymmetrical features, warped eyes, uneven pupils, malformed mouths, strange teeth, distorted ears, melted cheeks, inconsistent facial proportions, awkward expressions, and subtle identity changes between images. Objects should frequently fuse together, with warped shapes, incorrect proportions, strange geometry, ambiguous boundaries, duplicated features, missing details, incorrect perspective, and accidental shapes. Use soft blurry edges, smeared textures, muddy colors, blown-out highlights, crushed shadows, washed-out areas, noisy gradients, compression artifacts, and primitive
document fetched 2026-10-03 00:20 · canonical https://reddit.com/r/ChatGPT/comments/1wvy93f/gpt_image_25_craiyon