Your images stay on your device

This site is a static build: HTML, CSS and JavaScript served from a CDN. There is no application server and no upload endpoint. When you choose an image, the browser reads it into memory and hands it to a Web Worker that runs the parser, the cleaner and the encoder. Results, previews and downloads are created as in-memory objects in your tab. Closing the tab or removing an image from the list releases them.

What the site does not collect

  • No image files, file names or raw metadata values are uploaded.
  • No account, sign-in, cookies or local storage are used.
  • Image processing does not require uploading your files to a server.
  • Fonts are bundled with the site, so no third-party font request is made when a page loads.

Site analytics

The production site uses self-hosted Umami analytics to understand page visits and tool usage. Usage events can include file format, scan verdict, generator category and the action taken. Your images, file names and raw metadata values are not uploaded. Image processing happens locally. Analytics can be disabled in the site’s build configuration.

What a host may still see

Like any website, the static host that serves these pages may log ordinary request data such as IP address, time, requested path and browser type. Those logs relate to page loads only; image content never leaves your browser, so it cannot appear in them.

How cleaning is verified

The lossless path rebuilds the file from the original image data, color profile and a minimal orientation block, then parses the result again. The download button only appears when the re-check finds no removable metadata and the encoded pixel data matches the original byte for byte. The re-encode path inspects its output the same way. Every image check runs locally, without uploading the file.

What the site cannot verify

  • It reads C2PA manifests but does not validate signatures, certificates or trust lists.
  • It does not analyze pixels for AI generation and does not detect or remove invisible watermarks.
  • A verdict of “No AI label found” describes the file’s metadata, not how the picture was made.
  • Platform behavior described in the guides reflects public announcements and observed behavior as of the date on each guide, and platforms change their systems without notice.

Open standards used

Parsing follows the JPEG, PNG (third edition) and WebP (RFC 9649) container formats, the EXIF and XMP specifications, the IPTC Photo Metadata Standard’s digital source type vocabulary, and the C2PA specification for manifest stores in JUMBF boxes.

Contact

Questions about this page or the tools can be sent to hello@ailabel.app.