Video Watermark Remover

A command-line tool to remove watermarks from videos using OpenCV's inpainting algorithm.

Features

• Remove multiple watermarks from a single video
• Specify exact timing for each watermark appearance
• Precise coordinate-based watermark removal
• Multiple mask shapes: ellipse (default), circle, rectangle
• Feathered mask edges for smoother blending
• Configurable inpainting algorithm: TELEA (fast) or NS (higher quality)
• Multi-pass inpainting and adjustable radius for better results
• Progress tracking during processing
• Support for various video formats

Installation

1. Set up a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

2. Install dependencies:

   pip install -r requirements.txt

Usage

Basic Command Structure

python remover.py -i <input_video> -o <output_video> -w <watermark_definitions>

Watermark Definition Format

Each watermark definition accepts 6 to 11 comma-separated values:

• x,y,w,h,start_sec,end_sec[,shape[,feather[,algorithm[,radius[,passes]]]]]

Where:

• x, y: Top-left corner coordinates of the watermark
• w, h: Width and height of the watermark
• start_sec, end_sec: Time range when the watermark appears (in seconds)
• shape (optional): one of ellipse (default), circle, rect
• feather (optional): edge blur amount for mask (default: 5; 0 = sharp)
• algorithm (optional): telea (default) or ns (Navier-Stokes)
• radius (optional): inpainting radius in pixels (default: 7; typical 1–25)
• passes (optional): number of inpainting iterations (default: 1; 1–5)

Example Usage

For a video with three watermarks:

• Watermark 1: Appears from 0-3 seconds at position (50,50) with size 100x20
• Watermark 2: Appears from 3-6 seconds at position (300,400) with size 120x30
• Watermark 3: Appears from 7-10 seconds at position (500,100) with size 80x25

# Basic (default settings; ellipse shape, telea algorithm, 1 pass, radius 7)
python remover.py -i my_video.mp4 -o video_clean.mp4 -w "50,50,100,20,0,3" "300,400,120,30,3,6" "500,100,80,25,7,10"

# Rectangle mask with default feather, TELEA (fast)
python remover.py -i my_video.mp4 -o out_rect.mp4 -w "50,50,100,20,0,3,rect"

# Ellipse with heavy feathering and 2 passes, radius 10 (balanced quality)
python remover.py -i my_video.mp4 -o out_balanced.mp4 -w "50,50,100,20,0,3,ellipse,10,telea,10,2"

# Circle mask, NS algorithm, 3 passes, radius 12 (higher quality)
python remover.py -i my_video.mp4 -o out_quality.mp4 -w "300,400,120,120,3,6,circle,5,ns,12,3"

# Ensure output has no audio track (requires ffmpeg installed)
python remover.py -i my_video.mp4 -o out_muted.mp4 -w "50,50,100,20,0,3" --mute

Tips

• For simple logos: keep defaults or use telea,7,1
• For complex/gradient watermarks: try ns,10,2
• For large watermarks: try ns,12-15,2-3
• Increase feather (e.g., 8–12) to reduce visible edges

Finding Watermark Coordinates

To find the exact coordinates of watermarks in your video:

1. Use a video player like VLC to pause at the frame where the watermark appears
2. Note the pixel coordinates of the top-left corner
3. Measure the width and height of the watermark area
4. Note the time range when the watermark is visible

Alternatively, you can use OpenCV to create a simple coordinate finder script.

How It Works

The tool uses OpenCV's cv2.inpaint() to fill watermark regions based on surrounding pixels.

• Algorithms:
  • INPAINT_TELEA (default): fast and suitable for most use-cases
  • INPAINT_NS (Navier-Stokes): slower but can produce smoother results
• Quality controls:
  • Feathered masks soften edges before inpainting
  • Adjustable inpainting radius controls neighborhood size
  • Multi-pass inpainting (passes) can further refine the output

Requirements

• Python 3.7+
• OpenCV (opencv-python)
• NumPy

Output

The processed video will be saved with the same frame rate and dimensions as the input video, but with watermarks removed from the specified areas and time ranges.