Gemini Image Generation
Generate high-quality AI images using Google's Gemini model via OpenRouter API.
Capabilities
- Text-to-image: Create images from text descriptions
- Image-to-image: Transform existing images based on prompts (style transfer, editing, variations)
Requirements
Before using this skill, ensure:
-
Python packages are installed:
pip install openai python-dotenv -
API key is set:
export OPENROUTER_API_KEY="your-openrouter-api-key"Get your key at: https://openrouter.ai/keys
Usage
Text-to-Image
Generate an image from a text description:
python scripts/generate_image.py --prompt "A serene Japanese garden with cherry blossoms" --output garden.png
Image-to-Image
Transform an existing image:
python scripts/generate_image.py --input photo.jpg --prompt "Transform into a watercolor painting style" --output watercolor.png
Script Arguments
| Argument | Short | Required | Description |
|---|---|---|---|
--prompt | -p | Yes | Text description of the desired image |
--output | -o | No | Output file path (default: generated_image.png) |
--input | -i | No | Input image for image-to-image mode |
--model | -m | No | Model ID (default: google/gemini-3-pro-image-preview) |
Prompt Tips
For best results:
- Be specific: Include details about style, lighting, composition, and mood
- Reference art styles: "in the style of impressionism", "photorealistic", "anime style"
- Describe lighting: "soft morning light", "dramatic shadows", "golden hour"
- Include atmosphere: "mysterious", "cheerful", "melancholic"
Example Prompts
Portrait:
A portrait of an elderly fisherman with weathered skin and kind eyes,
wearing a knit cap, soft natural lighting, shallow depth of field,
photorealistic style
Landscape:
A misty mountain valley at dawn, pine forests in the foreground,
snow-capped peaks in the distance, rays of sunlight breaking through
clouds, epic cinematic composition
Abstract:
Abstract fluid art with deep ocean blues and metallic gold,
swirling patterns resembling galaxies, high contrast,
luxurious and elegant mood
Workflow
- Understand what the user wants to create
- Craft a detailed, descriptive prompt
- Run the generation script
- If the user provides an image, use image-to-image mode
- Present the generated image to the user
- Offer to iterate or adjust based on feedback
Troubleshooting
"OPENROUTER_API_KEY not set"
- Set the environment variable:
export OPENROUTER_API_KEY="your-key"
"openai package not installed"
- Install it:
pip install openai
"No images found in response"
- The model may have returned text instead of an image
- Try rephrasing the prompt to be more visual/descriptive
- Check if your API key has credits available
Output
The script will:
- Print progress messages
- Save the image to the specified output path
- Print "SUCCESS" with the file path on success
- Return exit code 0 on success, 1 on failure