Open methodology

How AnalyzeImage works

A transparent description of local file processing, parser boundaries, validation and evidence limits.

01

Local-first processing

For supported tools, the image file itself is read by browser code and Web Workers. AnalyzeImage does not provide an upload endpoint for these workflows. Standard requests for the site application still occur.

  • Magic-byte format detection
  • Header-level dimension checks
  • EXIF/XMP/IPTC normalization
  • Privacy-risk rules
  • Optional local SHA-256
02

Cleaning methodology

JPEG metadata segments and PNG/WebP metadata chunks are removed from a new copy. Pixel or compressed scan data is retained on the supported path. Orientation and ICC color information receive special treatment.

  • Original file remains unchanged
  • Output is reparsed by the interface
  • Claims are limited to tested formats and paths
03

Safety and session diagnostics

Every interactive tool uses the same stable error-code catalog and recovery guidance. A capped session log can be downloaded for troubleshooting, but it stays in the current browser tab and excludes image bytes, filenames and parsed metadata.

  • 40-entry session cap
  • No diagnostic upload endpoint
  • 30-second worker safety timeout
  • Format, file-size, pixel and batch limits
  • Per-file failure isolation
04

Evidence limits

Metadata is editable. Missing metadata does not prove safety or authenticity. C2PA validation documents signed provenance claims; it is not a general truth detector.

Put the model to work

Inspect an image without creating an upload.

Open metadata viewer