Your file, on your device.
Drop a file or try our example. Processing happens in your browser, without an upload.
Inspect AI provenance and hidden metadata. Choose what to remove
and download a clean copy—all in your browser.
Free to start. No account required. Files stay on your device.
Open a file to see what's behind it.
We show evidence, not guesses. Missing metadata doesn't prove a file is human-made. Proprietary watermarks are not checked.
From a hidden field to a ready-to-share copy,
in three straightforward steps.
Drop a file or try our example. Processing happens in your browser, without an upload.
Review embedded provenance, generator claims and personal information. No invented AI scores.
Remove supported metadata and download a new copy. Your original stays untouched.
Metadata can name an author or a tool, and can be edited. Content Credentials add a signature and a history of recorded actions. A valid signature doesn't independently prove every claim.
Some signals are encoded in pixels or language patterns. Those checks are not integrated here. Removing metadata does not establish that a file is free of watermarks.
Selective cleanup. Explicit limits.
No “remove everything” promises.
| Format | Inspection | Supported cleanup |
|---|---|---|
| Document properties and catalog XMP | Complete rewrite of selected properties/XMP in unsigned, unencrypted classic-xref PDFs; 10 MB / 100 pages | |
| PNG | Text, EXIF, XMP, credential containers | Selected text/EXIF fields and eligible XMP/credential containers; IDAT and color chunks preserved |
| JPEG | EXIF, XMP, comments, ICC, credential containers | Selected EXIF fields, comments and eligible XMP/credential containers; encoded scans and ICC preserved |
| WebP | EXIF, XMP, ICC, credential containers | Selected EXIF fields and eligible XMP/credential containers; encoded frames and alpha preserved |
| DOCX | Core, application and custom document properties | Selected properties only; other package entries preserved |
| UTF-8 text | Invisible characters and unusual spaces | Selected character classes; joining characters and emoji retained by default |
Image previews limited to 40 megapixels. EXIF rendering fields and XMP appearance settings are protected. Signed DOCX and unsupported structures are rejected. PDF: classic cross-reference files up to 10 MB and 100 pages; no encryption, signatures, forms or active content. C2PA signer trust is not configured. See detailed capabilities →
The same processing quality. The same privacy.
Pay only for a more convenient workflow.
USD. Paid jobs: 50 MB per file, 250 MB total. DOCX has additional archive safety limits. Purchase availability depends on merchant setup. Refund requests within 14 days; statutory rights remain unaffected.
No. It reports supported embedded evidence. Editable metadata can name an AI tool; a credential can record an AI action. Neither an empty result nor a signed record proves all aspects of a file’s origin.
Files opened here are processed locally. If you explicitly open a direct URL, your browser requests that file from its source. Account and payment requests are separate and contain no file data.
Supported image operations edit metadata containers without re-encoding image data. Color profiles, orientation and rendering fields are protected. The engine verifies preserved encoded bytes before offering a download.
Batch and folder processing, saved local presets, ZIP downloads and local delivery reports. Single-file inspection and cleanup stay free, with the same quality.
No. Both offers are one-time payments with no automatic renewal. Additional purchases extend your valid access.