watermarks-remover

Why we included this project

AI assistants from several vendors now embed provenance marks in the text and files they produce, which matters if you want to publish or share that content without the marks. This project strips them in layers: deterministic Python scripts scrub invisible Unicode, exotic spaces, and bidi characters from text; an optional rewrite hook targets token-sampling watermarks like SynthID-class marks; and file-level cleaning removes C2PA, EXIF, and XMP metadata from PNG, JPEG, PDF, DOCX, and HTML. It ships both as an agent skill for Grok-style environments and as standalone stdlib-only scripts, so the text and metadata layers run without any model calls. The project is explicit that it is meant for content you own, which makes it a privacy and hygiene tool rather than a way to launder someone else's work.