- category
- watermark-remover
- maturity
- growing
- approach
- CLI and Python library that strips visible AI-vendor labels (Gemini/Nano Banana sparkle, Doubao, Jimeng, Qwen, Kling, and others), uses diffusion-based regeneration to disrupt invisible pixel/frequency watermarks (e.g. SynthID), and strips AI-provenance metadata (C2PA, EXIF, XMP, IPTC, China TC260 AIGC labels).
- access
- open-source
- owns
- An Apache-2.0 CLI/Python/ComfyUI toolkit for removing AI watermarks and provenance metadata from images the user generated themselves; also offered as a paid, no-install hosted version.
- distinctFrom
- Unlike noai-watermark (a separate, independently-authored open-source project with similar but distinct functionality, 147 stars), this is a much larger project (4.3k stars) maintained by wiltodelta; it is also the open-source engine directly behind the already-tracked raiw hosted service (raiw.cc is explicitly listed in this repo's README as its no-install online version).
- notes
- Apache-2.0 license, 4.3k stars, 399 forks. PyPI package (remove-ai-watermarks) at version 0.20.1, released July 26, 2026 (within this run's window), confirming active current maintenance even though the project itself is older (HN Show HN post approx. May 2026). Legal disclaimer in the README: 'for lawful use on content you own... does not target stock agency previews or other watermarks that protect third party paid content.'
