convez.ai
Skill
DaYePython

image-to-prompt-skill

DaYePython/image-to-prompt-skill

Ver no GitHub
InicialPluginScore · de 0 a 100

Analyze images and reverse-engineer detailed prompts for image generation models using Claude's vision capabilities. Features quantitative complexity assessment, multi-image comparison, and model-specific exports.

Instalação

$npx convez-ai DaYePython/image-to-prompt-skill

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
1
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

README

A tradução para o português deste README ainda está na fila.

Image-to-Prompt Skill

License: MIT Claude Skills

Analyze images and reverse-engineer detailed prompts for image generation models using Claude's vision capabilities.

English | 中文

Overview

This skill uses Claude's vision capabilities to analyze any image and generate comprehensive prompts that could recreate it. Provides adaptive analysis depth based on image complexity with technical parameters, artistic context, and variation suggestions.

Key Features:

  • 🎯 Quantitative complexity assessment with clear formulas
  • 📊 Multi-image comparison with 8-dimension matrix analysis
  • 🎨 Three-tier analysis system (Quick/Detailed/Complete)
  • 🔄 Model-specific exports (Stable Diffusion, Midjourney, DALL-E)
  • 📚 Built-in style guides and reference materials

Installation

  1. Copy the image-to-prompt directory to ~/.agents/skills/
  2. Restart Claude Code or reload skills
  3. The skill will be available for use

Usage

Basic Invocation

Trigger the skill by:

  • "What prompt would create this image?"
  • "Reverse engineer this image into a prompt"
  • "Analyze this image and give me a prompt"
  • Attach an image and ask for prompt generation

Supported Formats

  • Primary: JPEG, PNG, WebP, BMP, TIFF
  • Animated (first frame): GIF, APNG

Output Tiers

Tier 1 - Quick Prompt (50-100 words)

  • Concise, essential elements only
  • Good for quick recreation attempts

Tier 2 - Detailed Prompt (200-300 words)

  • Structured by category (subject, style, lighting, etc.)
  • Includes negative prompts
  • High-fidelity recreation ready

Tier 3 - Complete Analysis (400-600 words)

  • Everything from Tier 2, plus:
  • Technical parameter recommendations
  • Artistic analysis and influences
  • 3-5 variation suggestions
  • Refinement tips

Export Options

  • Stable Diffusion: Prompt + Negative Prompt + Settings
  • Midjourney: Compressed format with parameters
  • DALL-E: Optimized natural language
  • JSON: Structured data export

Features

Core Capabilities

  • Quantitative Complexity Assessment - Formula-based tier selection for consistent results

    • Simple: 1 subject + solid/blurred background + no directional light
    • Moderate: 1-2 subjects + scene background + directional light
    • Complex: 3+ subjects + dramatic lighting + artistic treatment
  • Multi-Image Comparison - Compare multiple images with structured 8-dimension matrix:

    • Primary Subject, Style/Medium, Lighting, Color Palette
    • Composition, Mood, Complexity, Estimated Quality
    • Similarities, Differences, and Recommendations
  • Adaptive Analysis Tiers

    • Tier 1: Quick 50-100 word prompts
    • Tier 2: Detailed 200-300 word structured analysis
    • Tier 3: Complete 400-600 word analysis with technical parameters
  • Model-Specific Exports

    • Stable Diffusion format with positive/negative prompts and settings
    • Midjourney format with parameters (--ar, --style, --v)
    • DALL-E optimized natural language
    • Generic JSON for API integration
  • Additional Features

    • Interactive tier adjustment
    • Section extraction (lighting, technical params, etc.)
    • Variation exploration (3-5 alternative approaches)
    • Minimal information detection
    • Animated format handling (first frame analysis)

Examples

Quick Analysis (Tier 1)

User: [attaches portrait photo] What prompt created this?

Skill: A professional headshot of a person in business attire against a neutral 
gray background, shot with natural window light from the left creating soft shadows. 
Clean, sharp photography with shallow depth of field and warm color grading.

Multi-Image Comparison

User: [attaches 3 landscape photos] Compare these

Skill: [Generates comparison matrix across 8 dimensions]
       [Analyzes similarities: all use golden hour lighting, wide angles]
       [Notes differences: varying subjects, color palettes]
       [Recommendations for creating consistent set]

Documentation

  • SKILL.md - Complete skill implementation with all analysis logic
  • docs/spec.md - Design specification and architecture
  • docs/plan.md - Implementation plan and task breakdown
  • references/ - Style guides, technical parameters, examples

Contributing

Contributions welcome! Please read the design spec and implementation plan first.

License

MIT License - See LICENSE file for details.

Acknowledgments

Built with Claude's vision capabilities and skills framework.

Ver o original no GitHub

Bom para

Relacionados