Open-source research tool for inspecting, benchmarking, and scrubbing AI document metadata and invisible tracking artifacts.
-
Updated
Aug 21, 2026 - Python
Open-source research tool for inspecting, benchmarking, and scrubbing AI document metadata and invisible tracking artifacts.
Invisible, key-verifiable watermarks for AI/LLM-generated text — SynthID-Text-style, in Rust + WebAssembly (Node + browser). Live playground included.
Remove AI text watermarks (SynthID-Text, KGW, Unigram) by regenerating text from its meaning — scheme-blind, runs on local models, every claim benchmarked.
SynthID-Text: detector_mean on DeepMind public 30 keys (public-deepmind-30, ngram_len=5). Key-free twins, same prompt mixin on/off: 10/12 last-4. Not Claude. Not a remover.
Detect and remove hidden Unicode text watermarks locally, then verify LLM watermark mitigations with evidence-based tools and a CLI.
To associate your repository with the synthid-text topic, visit your repo's landing page and select "manage topics."