Drop files or click
PNG · JPG · WEBP · AVIF · HEIC · GIF · TIFF · BMP · PDF · DOCX · XLSX · PPTX · ODT · EPUB · SVG · HTML · MD · TXT
What it removes
Where the other tools erase the watermark you can see, this erases what is left behind. Files record their own origin, so an image whose logo you removed can still announce “made by AI” in its metadata.
- C2PA / Content Credentials — provenance manifests (JUMBF) embedded in images, PDFs and Office files
- XMP · EXIF — AI-generation markers such as
digitalSourceTypeandtrainedAlgorithmicMedia - Document authorship —
dc:creatorandApplicationfields in DOCX/PPTX/XLSX, plus C2PA inside embedded images - Invisible Unicode — zero-width spaces, tag characters and bidi controls hidden in text
What it cannot do
Plainly: this works on metadata and characters. It does not remove watermarks embedded statistically in pixels (SynthID and friends) or spread across word choice in text. Removing those means regenerating the image or rewriting every sentence — and no tool can honestly promise it will defeat an official detector.
Are my files uploaded?
No. The Python engine runs inside your browser via WebAssembly. After the one-time ~11 MB engine download, nothing leaves your device.
Does it degrade image quality?
No. Pixels are untouched — only metadata regions are cut out. JPEGs are not re-encoded; the scan data is preserved byte for byte.
Will it break emoji or non-Latin text?
No. Legitimately invisible characters — flag-emoji tags, skin-tone modifiers, Hangul jamo, Arabic and Mongolian joiners — are recognised and kept.
What is the engine?
The Python engine from the open-source watermarks-remover project (MIT), run unmodified in your browser.
Use only on content you own. It edits metadata and invisible characters only; you are responsible for how you use it. Not a tool for academic fraud.