What it does

This repo collects prompts for OpenAI’s GPT Image 2, organized by category: photorealism, game screenshots, UI mockups, typography, anime styles, and more. Each entry includes the full prompt text, a sample output image, and attribution to the X (Twitter) user who posted it. It’s essentially a community-curated cheat sheet for a specific image model.

The interesting bit

The prompts are deliberately verbose and oddly specific — “harsh direct on-camera flash,” “visible subtle skin texture and micro pores,” “no plastic skin, no digital over-sharpening.” The insight is that GPT Image 2 seems to need explicit negative constraints and camera-lens vocabulary to produce convincing fakes, not just descriptions of the scene.

Key highlights

• Side-by-side comparisons with other models (Nano Banana 2) in some entries
• Categories span photography, game UI, social-media posts, infographics, and character consistency
• Prompts sourced from named X creators with direct links to posts
• Available in eight languages via translated READMEs
• Links out to a web version at cyberbara.com

Caveats

• Heavy tilt toward photorealistic portraits with repetitive, occasionally uncomfortable descriptive patterns
• No explanation of why these prompts work — just the raw text and results
• Star count (1,558) suggests traction, but the repo itself is just markdown and images

Verdict

Useful if you’re actively prompting GPT Image 2 and want to skip the trial-and-error. Skip if you’re looking for general prompt-engineering theory or non-OpenAI models.