This repository offers a curated collection of reusable JSON prompt templates and style references for AI image generation, addressing the common issue of inconsistent and difficult-to-iterate AI prompts. It provides a structured, plug-and-play approach for users seeking to achieve more consistent visual styles across multiple generations with LLM-based image workflows.

How It Works

Each visual style is distilled into a self-contained style.json file, containing a prompt_template with variable placeholders, environment_variables declarations, example cases, and fidelity/avoidance parameters. Users copy this JSON into compatible LLM image generation tools, substitute values for the declared variables, and generate final prompts. This method enables consistent style direction and reusability, moving beyond ad-hoc text prompts.

Quick Start & Requirements

• Primary install / run command: Browse the "Featured Styles" or "All Styles" gallery, open a style folder, copy the style.json file, and paste it into a compatible LLM image workflow. Provide values for variables within the JSON or edit example cases.
• Non-default prerequisites and dependencies: Requires multimodal LLM image models capable of processing long, structured JSON prompts, such as ChatGPT Images 2 or Nano Banana Pro.
• Links:
  • All Styles gallery: [Link to All Styles gallery - implied by README structure]
  • Documentation/Catalog: docs/CATALOG.md [Link to docs/CATALOG.md - implied by README structure]

Highlighted Details

• Features a library of 68 distinct visual styles, categorized into themes like Photo + Doodle, Zine + Collage, Type Posters, Travel + City, and Editorial + Minimal.
• Each style.json is comprehensive, including prompt_template, environment_variables, examples, style_fidelity_anchors, source_content_to_avoid, and negative_prompt for fine-grained control.
• Provides a detailed "Complete Example" walkthrough demonstrating the input-to-output process for generating a specific poster style.

Maintenance & Community

The repository is curated by @VigoCreativeAI with assistance from OpenAI Codex, and the prompt "Star this repo to follow new style drops" suggests ongoing updates. No specific community channels (e.g., Discord, Slack) are mentioned.

Licensing & Compatibility

The repository structure and documentation are MIT licensed. Individual style.json files declare their own specific licenses, and preview images are provided as visual references. The MIT license generally permits commercial use and integration into closed-source projects.

Limitations & Caveats

Full galleries are not included within the repository itself. Users are instructed not to publish sensitive assets like original reference images, watermarks, or private prompts without permission.