Where to Find GPT-Image2 Prompt Examples Categorized by Visual Style

The GPT-Image2 prompt library in the freestylefly/awesome-gpt-image-2 repository stores all visual-style categories, scene tags, and ready-to-use templates in two key locations: the master JSON database at data/style-library.json and a human-readable markdown reference at agents/skills/gpt-image-2-style-library/references/style-library.md.

Finding well-organized GPT-Image2 prompt examples by visual style requires knowing exactly where the repository structures its design taxonomy. This guide covers the primary data sources, how to access them programmatically, and how to use the built-in agent skill for querying styles on demand.

---

The Master Style Library: data/style-library.json

The canonical source for all GPT-Image2 prompt examples categorized by visual style lives in data/style-library.json. This single JSON file defines the complete hierarchy of design options.

The file structure contains four top-level arrays:

• categories — High-level groupings like UI & Interfaces, Posters & Typography, and Illustration
• styles — Specific visual techniques such as 3D, Brand, Photography, and Minimal
• scenes — Contextual tags like Tech, Commerce, Nature, and Lifestyle
• templates — The actual prompt strings, each paired with style tags and usage guidance

Every template entry includes: title, description, category, styles array, scenes array, and a useWhen field explaining optimal use cases.

---

Human-Readable Reference: style-library.md

For browsing without parsing JSON, the repository auto-generates a formatted guide at agents/skills/gpt-image-2-style-library/references/style-library.md.

This markdown file groups all templates under their visual-style headings. Each entry displays:

• Title and description of the prompt
• Category, style tags, and scene tags
• "Use when" guidance for selecting the right prompt

This reference serves as the quickest way to scan GPT-Image2 prompt examples by visual style without writing code.

---

Accessing the Style Library in Your Application

Fetching Templates in a Web Frontend

The repository's own React frontend loads the JSON at runtime. The fetch call appears in src/main.jsx at line 3623:

// Example: fetch the style library and filter for UI-style templates
fetch('/style-library.json')
  .then(response => response.json())
  .then(data => {
    const uiTemplates = data.templates.filter(template =>
      template.styles.includes('UI')
    );
    console.log('UI-style prompt examples:', uiTemplates);
  });

This pattern works in any JavaScript environment with network access to the JSON file.

Programmatic Access via the NPM Package

The @freestylefly/gpt-image-2-style-library package provides typed access:

import { getStyleLibrary } from '@freestylefly/gpt-image-2-style-library';

(async () => {
  const library = await getStyleLibrary();  // Loads and parses the JSON
  
  const posterTemplates = library.templates.filter(template =>
    template.styles.includes('Poster')
  );
  
  console.table(posterTemplates.map(t => ({
    name: t.title,
    description: t.description,
    useWhen: t.useWhen
  })));
})();

---

Using the Agent Skill for Interactive Queries

The repository includes a Claude/Codex skill for conversational access to GPT-Image2 prompt examples.

Installation

# Install the skill globally

gpt-image-2-style-library install all

Querying by Visual Style

# Request prompts matching specific style criteria

gpt-image-2-style-library ask "Give me a 3D product render prompt for e-commerce"
gpt-image-2-style-library ask "Show minimal UI screenshot prompts for SaaS dashboards"

The skill parses natural language, matches against the styles and scenes arrays in the JSON, and returns ranked template suggestions.

---

Key File Reference

File Path	Purpose
data/style-library.json	Master database: categories, styles, scenes, templates
agents/skills/gpt-image-2-style-library/references/style-library.md	Browsable markdown grouped by visual style
src/main.jsx (line 3623)	Frontend implementation of runtime JSON fetching
agents/skills/gpt-image-2-style-library/SKILL.md	Agent skill installation and query syntax

---

Summary

• Primary source: The data/style-library.json file contains all GPT-Image2 prompt examples with complete metadata.
• Readable format: agents/skills/gpt-image-2-style-library/references/style-library.md presents the same data grouped by visual style.
• Three access methods: Direct JSON fetch, NPM package import, or agent skill conversation.
• Template structure includes searchable styles and scenes arrays for precise filtering.

---

Frequently Asked Questions

How do I filter prompts by multiple visual styles simultaneously?

Chain array methods on the templates array. The JSON structure supports multiple style tags per template, so templates.filter(t => t.styles.includes('3D') && t.styles.includes('Product')) returns prompts combining both aesthetics.

Can I contribute new prompt templates to the style library?

Yes. The repository accepts pull requests modifying data/style-library.json. Follow the existing schema: include title, description, category, styles array, scenes array, and a concise useWhen string explaining when to deploy this prompt.

Does the style library support versioning or API stability?

The JSON schema follows semantic versioning documented in SKILL.md. Breaking changes increment the major version, and the getStyleLibrary() function accepts an optional version parameter to pin against specific releases.

What's the performance impact of loading the full JSON?

The complete style-library.json is approximately 150KB uncompressed. For production applications, implement lazy loading or server-side filtering. The NPM package includes tree-shakable exports to import only specific category getters when bundle size matters.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}