GitHub - dinoBOLT/Gemini-Watermark-Remover: An AI powered extension to get rid of the Gemini watermark

3 min read Original article ↗

✨ Gemini Watermark Remover

License Version Platform

A professional, privacy-focused Chrome extension to remove watermarks from images generated by Google Gemini, running 100% locally in your browser.

This tool provides a clean, standalone interface to process your images. No data is ever sent to an external server, ensuring your privacy is fully protected.


🌟 Key Features

  • 100% Local & Private: All processing happens directly in your browser. Your images never leave your computer.
  • High-Quality Inpainting: Uses the powerful LaMa (Large Mask Inpainting) AI model to intelligently remove the watermark with minimal impact on the original image.
  • Preserves Original Quality: Only the small watermark area is modified. The rest of your image remains pixel-perfect and untouched.
  • Modern & Intuitive UI: A sleek, responsive interface with drag-and-drop support, a real-time progress bar, and detailed logs.
  • Before & After Comparison: An interactive slider lets you instantly compare the original and cleaned images to see the results.
  • Efficient Performance: The AI model is loaded once and cached for subsequent uses. A modular architecture ensures a smooth user experience.
  • Open Source: Fully transparent and open for community contributions.

🖼️ Screenshot

image

🚀 Installation

This extension is distributed as an unpacked extension and requires a manual installation process.

👉 Please read the INSTALLATION.md file for complete, step-by-step instructions on how to set up the project.

🛠️ How to Use

  1. Click the Gemini Watermark Remover icon in your Chrome toolbar. This will open the application in a new tab.
  2. Drag and drop an image file generated by Gemini into the designated area, or click to browse your files.
  3. The application will automatically start processing the image. You can monitor the progress via the progress bar and detailed logs.
  4. Once complete, the cleaned image will be displayed with a before/after comparison slider.
  5. Click the "Download Image" button to save the final, watermark-free image to your computer.
  6. To process another image, simply click "Process Another".

⚙️ Project Architecture

This project is built with a modern, modular JavaScript architecture to ensure maintainability and performance.

  • index.html: The main application interface.
  • styles.css: All styles for the user interface.
  • manifest.json: Defines the Chrome extension's properties and permissions.
  • background.js: A service worker that handles the extension's icon click.
  • src/: Contains all the core application logic.
    • lib/: The ONNX Runtime library and its required WASM files.
    • assets/: The lama_fp32.onnx AI model file.
    • js/: The modular JavaScript source code.
      • app.js: The main application entry point that orchestrates the entire process.
      • config.js: A centralized configuration file for all settings and constants.
      • ui-manager.js: Manages all DOM interactions, UI updates, and user feedback.
      • model-manager.js: Handles loading the ONNX model, running inference, and caching.
      • image-processor.js: Contains logic for image preprocessing, postprocessing, and final composition.
      • utils.js: A collection of helper functions for validation, logging, and more.

🔒 Privacy

Privacy is a core design principle of this project. The entire process, from loading the image to running the AI model and generating the final result, happens within a sandboxed environment in your browser. No data, images, or information is ever transmitted to any external server.

📜 Credits

This project is made possible by these incredible open-source technologies:

License

This project is licensed under the Apache License 2.0. See the LICENSE file for details.

🤝 Contributing

Contributions are welcome! Please feel free to open an issue or submit a pull request. See the CONTRIBUTING.md file for guidelines.