Wraps Google Gemini's image generation API with a structured editing layer that goes beyond basic text-to-image. You generate an image, decompose it into a JSON blueprint mapping every visual component (hair color, lighting type, clothing items), then edit specific fields using dot-notation paths like `subject[0].hair.color` or plain English instructions. Comes with ten asset presets for Facebook ads, hero images, OG cards, and YouTube thumbnails. Supports reference images for character consistency, dual model selection (Flash or Pro), and caches blueprints alongside generated files for quick re-edits. Useful when you need repeatable control over image variations without redoing entire prompts.
claude mcp add --transport stdio io.github.udhaykumarbala-gemini-image-studio uvx gemini-image-studio