Remove the visible Gemini AI watermark from images using reverse alpha blending. Use when asked to strip Gemini watermarks, batch-process Gemini images, or build/modify a CLI script that removes the bottom-right Gemini watermark without HTML or server-side components.
The Gemini Watermark Remover is a Python-based CLI tool that eliminates the visible Gemini AI watermark from images using reverse alpha blending techniques. When Google's Gemini AI generates images, it adds a semi-transparent watermark in the bottom-right corner. This skill provides a mathematical approach to remove that watermark by reversing the alpha compositing process, restoring the original pixel values without degrading image quality.
The tool features a straightforward command-line interface powered by Pillow for image processing. It includes pre-captured watermark alpha maps at different resolutions (48px and 96px) and implements detection logic to identify and remove the watermark automatically. The algorithm performs per-pixel calculations to reverse the blending operation, using the known watermark patterns stored in the assets folder. Output format flexibility is built-in, as Pillow automatically infers the image format from the output file extension.
This skill is ideal for content creators, designers, and developers who need to clean Gemini-generated images for professional use, batch process multiple watermarked images, or integrate watermark removal into automated workflows. It's particularly useful for those building media processing pipelines, preparing images for publication, or creating tools that need to handle Gemini AI outputs programmatically without manual editing.
Python 3.9 or higher is required. Install dependencies using pip install -r requirements.txt in the scripts folder.
The tool supports any image format that Pillow can read and write, including PNG, JPEG, WebP, and more. The output format is automatically determined by the file extension you provide.
The tool is designed for the standard Gemini watermark that appears in the bottom-right corner. Detection rules and limitations are documented in references/algorithm.md. Watermarks must match the pre-captured alpha maps.
The CLI processes one image at a time. For batch processing, you can write a wrapper script that calls remove_watermark.py in a loop or integrate the core algorithm into your own automation.
The mathematical approach, detection rules, and algorithm limitations are documented in references/algorithm.md. This file explains the reverse alpha blending process and edge cases.
Quick Setup:
.claude/skills/Repository
rookie-ricardo/erduo-skills