🧭 Overview

Watermark Remover is a cross-platform desktop application designed to remove static watermarks from video files. Its primary goal is to provide a private, local-first solution that avoids cloud uploads or subscriptions, allowing users to clean videos directly on their own hardware.

The project uses a modern “Hybrid” architecture, combining a web-based frontend with a powerful Python-based processing engine.

🔑 Key Features

• Interactive ROI Selection → Users can drag and resize a bounding box directly over the video frame to target the watermark using Konva.js.
• Multiple Removal Engines → Offers four distinct algorithms: Inpainting (TELEA), Gaussian Blur, Solid Fill, and Clone Stamp.
• Quick Preview → Generates a 3-second clip to test settings before committing to a full render.
• High Performance → Utilizes multi-core CPU processing and streams real-time progress from the backend.
• Audio Integrity → Preserves the original audio track by muxing it back into the output without re-encoding.
• Privacy-Focused → Operates entirely offline with no data leaving the machine and automatic cleanup of temporary files.

⚙️ How It Works

1. Input & Selection: The user loads a video into the Electron interface and draws a Region of Interest (ROI) over the watermark.
2. IPC Communication: The React frontend sends the coordinates and chosen removal method to the Python backend via Inter-Process Communication (IPC).
3. Frame Processing:
   • The Python backend uses FFmpeg to extract frames from the video.
   • OpenCV algorithms process each frame within the selected ROI using multiprocessing.Pool for speed.
4. Reassembly: FFmpeg muxes the original audio track back with the new frames to create the final video file.
5. Cleanup: All intermediate image files are deleted immediately upon completion.

🎬 Quick Start

To run the application locally:

# Clone the repository
git clone https://github.com/pandaind/watermark-remove.git
cd watermark-remove

# Install dependencies and start
npm install
npm run dev

📚 Topics Covered

• 🖥️ Desktop Shell → Electron 41 with React 19 and TypeScript
• 🎨 UI & Styling → Tailwind CSS v4 and Vite 8
• 🔧 Backend Processing → Python 3.11+ with Pydantic v2
• 📽️ Video Processing → FFmpeg 6 for extraction and reassembly
• 👁️ Computer Vision → OpenCV (cv2) and NumPy for inpainting algorithms
• 🏗️ Hybrid Architecture → Integrating Electron with a Python subprocess

🔧 Potential Extensions

Areas for experimentation:

1. Implement AI-based deep learning inpainting (e.g., LaMa)
2. Support for dynamic (moving) watermarks with object tracking
3. GPU-accelerated frame processing with CUDA
4. Batch processing of multiple videos
5. Cloud-sync for project configurations (optional)

🤝 Contributing

This project is an open-source tool. You can:

• Study the hybrid Electron-Python architecture
• Add new inpainting algorithms or filters
• Improve the performance of frame processing
• Submit bug reports or feature requests via GitHub