image-generation

Important (December 2025): This skill now uses the google-genai SDK. The google-generativeai package has been deprecated. See migration guide.

AI-powered image generation and editing using Google Gemini, Imagen, and OpenAI models.

Overview

This skill enables Claude Code to generate and edit images using state-of-the-art AI models from Google (Gemini) and OpenAI (DALL-E). Create photorealistic images, illustrations, logos, product mockups, and artistic compositions from natural language descriptions. Edit existing images with text instructions, apply style transfers, and refine outputs through iterative conversation.

Attribution: This skill is inspired by the gemini-imagegen skill from Every Marketplace by Every Inc. We're grateful for their pioneering work in AI image generation skills.

Quick Start

Installation

# Install the skill from the skillz repository
skillz install image-generation

# Or copy to your project
cp -r skills/creative/image-generation .claude/skills/

Requirements

API Keys:

# Set environment variables
export GEMINI_API_KEY="your-gemini-api-key"
export OPENAI_API_KEY="your-openai-api-key"

Python Dependencies:

pip install google-genai openai pillow requests

Basic Usage

Once installed, simply ask Claude to generate or edit images:

"Generate a professional headshot for LinkedIn"
"Create a logo for a coffee shop called Mountain Peak Coffee"
"Change the background of this photo to a beach sunset"
"Make this landscape look like it was painted by Van Gogh"

Claude will automatically:

1. Recognize the image generation request
2. Select the appropriate model
3. Craft an effective prompt
4. Generate the image
5. Save it and show you the result

Available Models

Google Gemini

• gemini-2.5-flash-image - Fast, 1024px, ideal for rapid iteration and editing
• gemini-3-pro-image-preview - Up to 4K, professional quality, complex compositions, 14 reference images

Google Imagen 4

• imagen-4.0-fast-generate-001 - Cheapest ($0.02/image), rapid generation
• imagen-4.0-generate-001 - High-quality photorealistic, excellent text rendering
• imagen-4.0-ultra-generate-001 - Highest quality (single image per request)

OpenAI

• gpt-image-1.5 (Recommended) - Fast, affordable, excellent editing & text rendering
• gpt-image-1 - Maximum resolution (4096x4096)
• dall-e-3 - Creative interpretations, automatic prompt expansion
• dall-e-2 - Budget-friendly, supports variations

Capabilities

Text-to-Image Generation

Generate images from natural language descriptions:

• Photorealistic photography
• Illustrations and digital art
• Logos and graphic design
• Product photography
• Character designs
• Landscapes and environments
• Abstract and artistic compositions

Image Editing

Modify existing images with text instructions:

• Background replacement
• Style transfers
• Adding or removing objects
• Lighting and mood changes
• Color grading and filters
• Perspective transformations

Iterative Refinement

Refine images through conversation:

• Multi-turn improvements
• Progressive edits
• Exploration of variations
• Style and composition adjustments

Documentation

Core Files

• SKILL.md - Complete skill implementation with detailed instructions
• README.md - This file, overview and quick start

References

• references/gemini-models.md - Complete Gemini API documentation
• references/dalle-models.md - Complete DALL-E API documentation
• references/prompt-engineering.md - Comprehensive prompt writing guide

Examples

• examples/text-to-image.md - 10 detailed text-to-image examples
• examples/image-editing.md - 10 image editing examples

Use Cases

Business & Marketing

• Product photography for e-commerce
• Marketing visuals and social media graphics
• Brand logos and identity design
• Advertising concept art
• Presentation graphics

Creative & Artistic

• Character design and concept art
• Book and magazine illustrations
• Album covers and poster designs
• Fantasy and sci-fi artwork
• Abstract and experimental art

Professional

• Architectural visualizations
• Interior design mockups
• Professional headshots
• Data visualization concepts
• Educational illustrations

Personal

• Social media content
• Creative projects
• Gift designs
• Personal branding
• Hobby and craft references

Example Prompts

Photorealistic Portrait

Professional business headshot, confident expression, neutral gray background,
soft studio lighting, shot on 85mm lens, shallow depth of field, high quality

Illustration

Cute robot character, kawaii style, pastel colors, bold outlines, cel-shading,
cheerful expression, sticker format with white border, children's illustration

Logo Design

Tech startup logo, minimalist geometric design, abstract network nodes, blue
and silver gradient, professional, vector style, clean modern aesthetic

Product Photography

Wireless earbuds with charging case, professional product photography, white
seamless background, studio lighting, 3/4 angle view, clean composition,
commercial e-commerce quality

Landscape

Mountain valley at sunset, mist over alpine lake, pine trees, warm golden light,
peaceful atmosphere, professional landscape photography, 24mm wide angle, high detail

Model Selection Guide

Choose gpt-image-1.5 (Recommended) when:

• Need excellent text rendering
• Iterative editing workflows (preserves faces/logos)
• Fast turnaround with good quality
• Balanced cost/quality

Choose Gemini Flash when:

• Multi-turn editing via chat
• Rapid iteration on concepts
• Speed is priority

Choose Gemini Pro when:

• Need 4K resolution
• Complex compositions with reference images
• Real-time data visualization (Search grounding)

Choose Imagen 4 Fast when:

• Budget is primary concern ($0.02/image)
• High-volume generation
• Text-to-image only (no editing)

Choose Imagen 4/Ultra when:

• Photorealistic quality
• Excellent text in images
• No editing needed

Choose DALL-E 3 when:

• Creative interpretation preferred
• Artistic/illustrative content

Choose gpt-image-1 when:

• Maximum resolution (4096x4096) needed

Tips for Best Results

1. Be Specific

Detailed prompts produce better, more consistent results.

Vague: "A nice landscape" Specific: "Mountain valley at sunrise, mist over lake, pine trees, warm golden light, peaceful"

2. Include Style Information

Always specify the artistic style or photographic approach.

Examples: photorealistic, digital illustration, oil painting, 3D render, vintage photography

3. Describe Lighting

Lighting dramatically affects mood and quality.

Examples: golden hour lighting, studio lighting, dramatic side lighting, soft diffused light

4. Specify Composition

Guide how elements should be arranged.

Examples: centered composition, rule of thirds, close-up, wide angle, overhead view

5. Iterate Thoughtfully

Use conversation to refine results progressively rather than starting over.

Troubleshooting

Image Quality Issues

• Use Gemini Pro or DALL-E 3 for higher quality
• Add technical details: "high resolution", "professional quality", "sharp focus"
• Be more specific about lighting and style

Unexpected Results

• Make prompts more explicit and detailed
• Specify what you don't want (via positive alternatives)
• Try a different model

Text in Images

• Use Gemini Pro (best for text rendering)
• Specify exact text in quotes
• Keep text simple and prominent

API Errors

• Verify API keys are set correctly
• Check rate limits and quotas
• Review content policy compliance
• Implement retry logic with backoff

Cost Optimization

1. Prototype with cheaper models (Gemini Flash, DALL-E 2)
2. Generate finals with premium models (Gemini Pro, DALL-E 3)
3. Use appropriate resolutions (don't use 4K when 1K suffices)
4. Batch similar requests together
5. Save successful prompts for reuse

Content Policies

Both Google and OpenAI have content policies that prohibit:

• Violence, gore, or disturbing content
• Hate symbols or discriminatory content
• Sexual or adult content
• Illegal activities
• Real identifiable people without consent
• Copyrighted characters or IP

Keep prompts appropriate and focused on original creative content.

Attribution & Credits

Inspiration

This skill was inspired by the gemini-imagegen skill from Every Inc.'s marketplace:

• Repository: https://github.com/EveryInc/every-marketplace
• Path: plugins/compounding-engineering/skills/gemini-imagegen
• Credit: Every Inc. for the original Gemini image generation implementation

Extensions

This implementation extends the original concept by:

• Adding OpenAI DALL-E model support (DALL-E 2 and DALL-E 3)
• Comprehensive documentation for prompt engineering
• Detailed examples for various use cases
• Model selection guidance
• Image editing workflows

License

This skill follows the skillz repository license. The original gemini-imagegen concept belongs to Every Inc.

Resources

API Documentation

• Google GenAI SDK Migration Guide
• Google Gemini Image Generation
• Google Imagen API
• OpenAI Images API

Pricing

• Google AI Studio Pricing
• OpenAI Pricing

Community

• Skillz Repository
• Every Marketplace

Contributing

Found ways to improve this skill? Contributions welcome:

1. Additional model support
2. More example use cases
3. Prompt engineering techniques
4. Error handling improvements
5. Documentation enhancements

Version History

• v1.2.0 (2025-12-29) - SDK Migration

  • Migrated from deprecated google-generativeai to google-genai SDK
  • Updated all code examples to new API patterns
  • Added migration guide references

• v1.1.0 (2025-12-XX) - New Models

  • Added Google Imagen 4 family (Fast, Standard, Ultra)
  • Added OpenAI gpt-image-1.5 and gpt-image-1
  • Updated model selection guidance
  • Added pricing information

• v1.0.0 (2025-01-23) - Initial release

  • Gemini 2.5 Flash and 3 Pro support
  • DALL-E 2 and 3 support
  • Comprehensive documentation
  • Text-to-image and image editing examples
  • Prompt engineering guide

Support

For issues or questions:

1. Check the examples directory
2. Review prompt engineering guide
3. Consult model-specific documentation (Gemini, DALL-E)
4. Open an issue in the skillz repository

---

Happy generating! 🎨✨

Built with inspiration from Every Inc.'s gemini-imagegen skill