gemini-image-gen

Text-to-image generation with Gemini converts descriptive prompts into high-quality images. The Skill accepts natural language descriptions and outputs images at configurable aspect ratios, saved to ./docs/assets. It interfaces with Gemini-2.5-flash-image model via the google-genai SDK.

Image editing and multi-image composition are supported features. You can modify existing images with editing prompts and combine up to three input images into a single scene. Pillow handles image processing while the Gemini model performs the creative transformations.

Configurable safety settings let you control content filtering on generated images. Watermarking is optional, and all outputs save automatically to ./docs/assets. The Skill returns images plus optional text responses from the model for iterative refinement.

Iterative refinement through conversation enables you to request adjustments to generated images by providing follow-up prompts. The Skill maintains context across requests, allowing multi-turn interactions that progressively improve results without regenerating from scratch.

The Skill accepts text prompts and image files as inputs, supporting flexible aspect ratios for output generation. Image inputs work with standard formats handled by Pillow; the Gemini API processes both text and image data together for composition and editing tasks.

Helper scripts abstract API complexity, but the Skill uses the Python SDK and requires basic familiarity with google-genai, requests, and Pillow dependencies. The Core Task Description indicates Advanced implementation depth, reflecting flexible customization capabilities rather than entry difficulty.