--- name: gemini-image description: Reference guide for using google-genai Python library to generate images with gemini-3-pro-image-preview model. Use this skill when building new projects that need Gemini image generation capabilities, to understand the correct API patterns, configuration options, and best practices. --- # Gemini Image Generation Guide Reference for generating images with Google's `gemini-3-pro-image-preview` model. ## Language References Load the appropriate reference based on the project's language: | Language | Reference File | |----------|------------------------| | Python | `references/python.md` | **Instructions:** When implementing Gemini image generation, read the corresponding language reference file for complete code patterns and examples. --- ## Model Information | Property | Value | |----------------------|------------------------------| | Model ID | `gemini-3-pro-image-preview` | | Cost | ~$0.134 per image (2K) | | Max Reference Images | 5+ (high fidelity) | | Resolutions | 1K, 2K, 4K | ## Supported Aspect Ratios | Ratio | Use Case | |--------|----------------------------| | `1:1` | Square, social media posts | | `2:3` | Portrait photos | | `3:2` | Landscape photos | | `3:4` | Portrait, mobile screens | | `4:3` | Standard display | | `4:5` | Instagram portrait | | `5:4` | Large format | | `9:16` | Vertical video, stories | | `16:9` | Widescreen, presentations | | `21:9` | Ultra-wide, cinematic | ## Image Sizes | Size | Resolution | Use Case | |------|------------|-------------------------------| | `1K` | ~1024px | Thumbnails, previews | | `2K` | ~2048px | Standard output (recommended) | | `4K` | ~4096px | High-quality prints | **Important:** Use uppercase "K" (not "1k", "2k", "4k"). --- ## Environment Setup ```bash export GOOGLE_API_KEY='your-api-key-here' ``` --- ## Core Capabilities ### 1. Text-to-Image Generation Generate images from text descriptions with configurable aspect ratio and resolution. ### 2. Style Transfer with Reference Images Pass reference images to maintain consistent style across generations. Supports up to 5+ images for high fidelity. ### 3. Image Editing Modify existing images based on text instructions (add/remove elements, style changes). ### 4. Batch Generation Generate multiple style candidates or variations. --- ## Prompt Engineering Tips ### Be Descriptive ``` Bad: "cat, sunset" Good: "A fluffy orange tabby cat sitting on a wooden fence, watching a vibrant sunset over rolling hills. Warm golden and pink light illuminates the scene. Photorealistic style with soft focus background." ``` ### Specify Visual Elements - **Lighting:** "soft morning light", "dramatic side lighting", "golden hour" - **Style:** "oil painting", "watercolor", "3D render", "photorealistic" - **Mood:** "serene", "dramatic", "whimsical", "mysterious" - **Composition:** "close-up portrait", "wide landscape", "bird's eye view" - **Camera:** "35mm lens", "shallow depth of field", "wide angle" ### For Style Transfer When using reference images, be explicit about what to transfer: - "Match the color palette and brushstroke style of the reference" - "Keep the artistic mood and lighting from the reference image" --- ## Common Issues | Issue | Solution | |----------------------|-------------------------------------------------------------| | "No image generated" | Check prompt for content policy violations; simplify prompt | | "Invalid image_size" | Use uppercase: `"1K"`, `"2K"`, `"4K"` | | "API key not found" | Set `GOOGLE_API_KEY` environment variable | | Rate limits | Add delays between requests; use exponential backoff | --- ## Pricing Comparison | Model | Cost per Image | |----------------------------|----------------| | gemini-3-pro-image-preview | ~$0.134 | | gemini-2.5-flash-image | ~$0.039 |