The prompt cookbook OpenAI forgot to ship

What it does This repo collects highly specific, often paragraph-long prompts for OpenAI’s gpt-image-2 model. The entries cover portraits, posters, UI mockups, game screenshots, and typography—each paired with the exact API call pattern so you can copy, paste, and generate.

The interesting bit The prompts read like obsessive art-direction briefs rather than typical diffusion-model tag soup. One portrait prompt spends fifty words specifying skin undertones, eyelid shape, and convenience-store fluorescent reflections. The collection treats prompt engineering as a craft, not a guessing game.

Key highlights

• Organized by real-world categories: portraits, posters, UI/UX mockups, character sheets, infographics, and more
• Every prompt includes a credited source, usually from X/Twitter users who originated the style
• Includes a minimal Python snippet showing the exact gpt-image-2 API call
• Heavy focus on photorealistic and film-stock aesthetics—35mm grain, CCD flash, “RAW iPhone quality”
• Promotes an alternative API provider (MuAPI) for users without OpenAI accounts

Caveats

• The repository is essentially a markdown list; there is no code, tooling, or automation beyond the single API example
• Several prompts lean heavily on specific physical and ethnic descriptors, which may not age well or suit all use cases
• The README badges reference a different repository (Awesome-GPT-Store), suggesting the page may be templated or inconsistently maintained

Verdict Grab this if you are building an image-generation app or just want to shortcut weeks of prompt iteration. Skip it if you are looking for a framework, model weights, or anything you can pip install.