guillaumemeyer/watermarks-remover

What it solves

This project provides tools to remove AI-generated provenance marks and watermarks from text and files. It is designed for users who own their content and want to ensure privacy and hygiene by stripping invisible markers used by AI vendors to identify machine-generated content.

How it works

The system operates across three primary layers:

• Layer A (Deterministic): Uses Python scripts to remove invisible Unicode characters, exotic spaces, bidirectional overrides, and tag characters.
• Layer B (Statistical): Addresses token-sampling watermarks through agent-driven rewrites or a specific rewrite hook.
• Files: Strips metadata and provenance marks from a wide variety of formats, including images (PNG, JPEG, WebP, etc.), documents (PDF, DOCX, XLSX, PPTX), and media files (MP4, WAV, MP3).

It can be deployed as a standalone HTTP service, a set of CLI scripts, or as a "skill" for AI agents (like Claude Code, Codex, or Cursor) that allows the agent to clean files it has written.

Who it’s for

• Users of AI agents who want to automatically clean files written by those agents.
• Content creators who want to remove AI provenance markers from their own documents and images.
• Developers looking to integrate AI watermark removal into their own applications via an HTTP API.

Highlights

• Multi-Vendor Support: Targets marks from Claude, Gemini (SynthID-Text), OpenAI, and open-LLM schemes (Kirchenbauer, Aaronson).
• Agent Integration: Includes specialized installers for Claude Code, Codex, Cursor, and Grok.
• Detailed File Support: Handles a massive range of formats from simple text to complex containers like PDF and DOCX.
• Deterministic Hooks: Can be configured as a PostToolUse hook to automatically check or clean files immediately after an agent writes them.