Gemini Watermark Remove Skill

繁體中文

Remove watermarks from Google Gemini AI generated images (Nano Banana).

✨ Features

• 🚫 Automatically remove Gemini watermarks
• 🔒 Fully local processing, protecting privacy
• ⚡ Batch processing support
• 🎯 Auto-detect image size and select appropriate mask

📦 Installation

Method 1: npx add-skill (Recommended)

Use add-skill for quick installation to various AI agents.

Auto-detect installed agents

npx add-skill kevintsai1202/GeminiWatermarkRemoveSkill

Install to specific agent

# Install to Antigravity
npx add-skill kevintsai1202/GeminiWatermarkRemoveSkill -a antigravity -g -y

# Install to Claude Code
npx add-skill kevintsai1202/GeminiWatermarkRemoveSkill -a claude-code -g -y

# Install to Cursor
npx add-skill kevintsai1202/GeminiWatermarkRemoveSkill -a cursor -g -y

# Install to Roo Code
npx add-skill kevintsai1202/GeminiWatermarkRemoveSkill -a roo -g -y

# Install to Gemini CLI
npx add-skill kevintsai1202/GeminiWatermarkRemoveSkill -a gemini-cli -g -y

Parameters:

• -a, --agent <agents...>: Specify target agent
• -g, --global: Install to global directory
• -y, --yes: Skip confirmation prompt

Method 2: Git Clone

# Clone to Antigravity skills directory
git clone https://github.com/kevintsai1202/GeminiWatermarkRemoveSkill.git ~/.gemini/antigravity/skills/gemini-watermark-remove

# Clone to Claude Code skills directory
git clone https://github.com/kevintsai1202/GeminiWatermarkRemoveSkill.git ~/.claude/skills/gemini-watermark-remove

Method 3: Manual Installation

1. Download ZIP or clone this repository
2. Copy the folder to the corresponding skills directory:
   • Antigravity: ~/.gemini/antigravity/skills/
   • Claude Code: ~/.claude/skills/

Verify Installation

Restart your AI agent or reload skills, then invoke through AI conversation.

🚀 Usage

After installing this Skill, invoke it through AI conversation:

Single Image

"Remove the watermark from this image: D:\images\photo.png"

Batch Processing

"Batch remove watermarks from all images in D:\images folder"

🛠️ Technical Principle

Uses reverse Alpha blending algorithm to precisely restore pixels covered by watermarks:

Original Pixel = (Current Pixel - α × Watermark Color) / (1 - α)

Auto-select mask based on image size:

• ≤1024px: Use 48×48 mask
• >1024px: Use 96×96 mask

📄 License

MIT License