What it does

This repo collects 100+ image-generation prompts and their outputs, mostly scraped from Twitter/X and the Sora community. Each entry pairs a screenshot with the exact Chinese prompt that produced it, so you can copy, tweak, and rerun it in ChatGPT or Sora.

The interesting bit

The value isn’t the images—it’s the prompt engineering archaeology. The curators translated and cleaned up viral X posts into reusable templates with bracketed variables (e.g., replace [真实物体] with “coffee bean”). It’s essentially a pattern library for GPT-4o’s native image generation and the newer gpt-image-1 model.

Key highlights

• 100 numbered cases spanning styles: Ghibli, 3D chibi, pixel art, glassmorphism, tilt-shift cyberpunk, faux-tweet screenshots
• Dedicated 40-case section for OpenAI’s official gpt-image-1 examples, translated to Chinese
• Prompts include attribution links back to original X posts
• CC BY 4.0 license on the curated collection itself
• Bilingual README (Chinese primary, English secondary)

Caveats

• Images are hosted in-repo and served raw from GitHub; no CDN or gallery UI
• No code, no CLI, no automation—purely a markdown catalog
• Prompt quality varies; some are elaborate narrative descriptions, others are sparse style tags

Verdict

Grab it if you’re hunting for proven GPT-4o prompt patterns and want to skip the trial-and-error phase. Skip it if you need a tool, API wrapper, or systematic evaluation of what actually works.