Remove the trace.Keep the work.
Strip invisible Unicode watermarks, SynthID-class statistical marks, document metadata, and image EXIF/XMP/C2PA — from content you own. Processed in memory, never stored.
Register idle — awaiting your first specimen
Three benches, one sheet
Everything runs in memory, on one specimen sheet — scan, clean, and download. Nothing is stored.
Text bench
Scan pasted text or a .txt / .md / .docx / .pdf file for invisible watermarks and AI traces, then clean.
- Invisible Unicode removal
- SynthID / KGW-class statistical scan
- AI-vs-human score, before & after
- DeepSeek rewrite — humanize, Layer B, MASH
Image bench
Strip EXIF / XMP / IPTC / C2PA metadata from .jpg, .png and .webp, then inpaint a visible watermark.
- EXIF / XMP / IPTC / C2PA removal
- Auto-detect the watermark region
- Diffusion inpainting
Detection report
A multi-detector ensemble readout — statistical, perplexity, neural, SynthID — with before/after percentages.
- Weighted ensemble score
- Per-detector breakdown
- SynthID z-score
- PDF report download
How it works
Scan
Paste text or drop a file. The register reads the trace — invisible characters, statistical marks, and metadata.
Clean
Strip the marks and metadata. Optionally rewrite the text with DeepSeek to break statistical patterns.
Download
Take the cleaned text, file, or image. Nothing was stored — only your output leaves the bench.
Legal in the UK. Yours when you pay.
You should be able to remove watermarks and metadata from content you own or are licensed to use — and keep full control of what you produce.
Lawful in the UK
Removing AI watermarks and metadata from content you own or have rights to is legal in the United Kingdom.
You own the output
When you pay for a service, the outputs belong to you. Your work is not used to train models, resold, or shared.
Nothing is stored
Files are processed in memory and discarded after processing. No copies, no archive, no data lake.
Not legal advice. Intended for content you own or have the rights to.
Choose your bench
Every plan processes files in memory and never stores them. Cancel anytime from the customer portal.
FAQ
Is removing AI watermarks legal?
In the United Kingdom, removing watermarks and metadata from content you own or are licensed to use is lawful. This service is intended for your own content — not for content you don't have rights to.
Do you store my files?
No. Files are processed in memory and discarded immediately after processing. Nothing is written to disk, archived, or used for training.
Who owns the output?
You do. When you pay for the service, the cleaned output belongs to you. We don't resell, share, or train on your work.
What is a SynthID watermark?
A statistical watermark embedded in token selection by some models, rather than visible characters. We scan for the class of statistical bias it leaves behind.
Will it pass every AI detector?
Honestly, no single tool guarantees that. Detection is an arms race. Our multi-detector ensemble and rewrite modes improve your odds, but we won't overpromise.
Can I cancel anytime?
Yes. Manage or cancel your subscription from the Stripe customer portal — your access stays active until the end of the current billing period.