Free Metadata Comparison

Metadata differences come first

The comparison uses available metadata collected by ExifTool. It retains grouped tags and repeated instances, including document properties, author and software fields, dates, camera settings, GPS, XMP, IPTC and other records that the format contains.

The table starts with changed or unknown embedded metadata. Turn off that filter to show matching fields, or search for a name or value. Readable values and numeric parser values are compared separately. A readable value can differ while its numeric value matches.

Browser filenames and last-modified times, container properties, derived values and parser information have their own categories. They do not inflate the embedded-metadata change count. A browser timestamp is not an original disk creation time.

If extraction is partial or a parser value is unavailable, the report says so. An absent field in incomplete results is unknown; it is not evidence that someone removed it. A complete report still describes only the metadata this parser collected.

Content and file bytes are separate checks

• PDF: compare native text extracted in page order. There is no OCR. Scanned pages, images, layout, signatures, annotations, attachments and other non-text elements are not fully compared.
• DOCX: compare selected main-document text, including table paragraphs. Headers, footers, comments, tracked-change meaning, embedded objects and formatting are outside this text comparison.
• XLSX: compare selected cell records, shared strings, formulas and stored values. This does not recalculate formulas or compare formatting, charts, drawings or external data.
• PPTX: compare selected slide text. Notes, animations, images, layout and embedded objects are outside this text comparison.
• JPEG, PNG and supported WebP: compare browser-decoded pixels within the image limits. Browser color conversion and orientation can affect the result. This is not a complete animation or multi-frame comparison.
• Other formats ExifTool recognizes: metadata and selected file bytes are still compared. Unsupported content extraction is marked unavailable, including formats the browser cannot decode.

Matching extracted text or decoded pixels does not mean that all file content is identical. Exact byte equality is reported separately. A changed byte count describes differing positions, not a claim about how many edits were made.

Your copies stay on your computer

Your selected files, labels, metadata and extracted content are not uploaded to GDF or an outside analysis service. The page, parsers, fonts and code download to your browser; inspection and comparison run locally.

This tool does not load optional analytics or advertising. It does not query AI services, DNS, location services or remote document converters. No account or email address is needed.

Clear removes the displayed comparison and stops its workers. It does not delete your original files. Downloaded reports remain wherever you save them and may contain private information.

Save the comparison

Download JSON, CSV, HTML or a paginated PDF. Display filters do not filter the downloads. JSON and HTML retain the collected inspection records and supported extracted text. CSV and PDF include the comparison table, input inventory, hashes, settings, versions, coverage, warnings and differences.

CSV uses reversible json: cell encoding to retain types, repeated fields and formula-like strings. Original structured values remain in JSON. This does not promise spreadsheet protection after decoding or editing.

Using the comparison in an examination

Work on copies and preserve the originals with their collection context. Each input receives a SHA-256 hash so you can document exactly which bytes were compared. Inspection time comes from the browser clock and is not independently verified.

Changes in metadata do not identify who made them, establish when they occurred or prove tampering. Matching metadata does not prove that two copies have the same content. Matching extracted text does not establish authenticity or digital-signature trust.

The tool uses the same extraction settings for each copy. It does not decrypt password-protected files, recover deleted records or validate certificates. Check material findings and limitations before relying on them in a report.

Browser limits

Two to four files, up to 20 MiB each. Metadata keeps its existing 60-second parser-loading, 15-second inspection and 4 MiB engine-output limits. Content has a separate 25-second deadline and bounded page, pixel, XML and text limits, recorded with each file. These deadlines are not a promise of a fixed browser-memory cap.

The comparison stops rather than silently discard records above 25,000 combined collected metadata records. The table displays 100 rows per page. Text changes show up to 2,000 lines on screen; the collected bounded text remains in JSON. Large changes use a clearly labeled changed-block display when line alignment would exceed the comparison limit.

PDF generation stops instead of returning a shortened report if it exceeds 25,000 export records, 4,194,304 represented UTF-16 code units, 500 pages, 25 MiB output or 60 seconds. Unsupported glyphs, right-to-left text and hidden control characters use visible Unicode escapes. Use JSON for the original values.