VigoZhao/AI-Visual-Prompt-Cookbook

Curated collection of reusable JSON prompt templates & style references for AI image generation. Updated daily.

What it solves

It addresses the difficulty of reusing, comparing, and iterating on AI image prompts, which are typically one-off text blobs. By distilling visual styles into structured formats, it enables more consistent and reusable style direction across different image generations.

How it works

Each visual style is stored as a style.json file. Users copy this JSON into a multimodal LLM (such as ChatGPT or Gemini 3 Pro Image) and provide specific values for defined variables (like subject, location, or text). The LLM then uses the prompt_template within the JSON to generate a final, detailed image prompt that maintains the visual system's fidelity.

Who it’s for

Digital artists, designers, and AI image creators who want to achieve professional, consistent visual styles (such as editorial posters, zines, or travel collages) without writing complex prompts from scratch.

Highlights

  • Structured Prompting: Uses JSON templates instead of plain text to ensure style consistency.
  • Extensive Library: Includes 104 curated visual styles across categories like Photo + Doodle, Zine + Collage, and Editorial + Minimal.
  • Variable-Based Customization: Allows users to easily swap subjects and text while keeping the overall art direction.
  • Multimodal Optimization: Specifically designed for models that can reason over structured data to produce high-fidelity images.