ExifTool, MAT2 or hexscope: what each leaves in a file

1 What was tested

Each tool in its default mode: exiftool -all= (ExifTool 13.55), mat2 (MAT2 0.15.0) and hexscope clean (0.5.0). The files are hexscope's test files: some written by real programs and devices — an iPhone, an Android phone, Keynote, Chrome — and many made for the tests. Each copy was read by ExifTool, which counts a tag only if it still holds something, and by hexscope, which also reads a PDF's earlier versions and the text hidden on its pages.

hexscope wrote this comparison, so it was built to be checked: the script, the files and every result are in the repository, and anyone can run it again.

2 Photos

• JPEG, PNG, AVIF: all three removed the location, the camera and the dates. MAT2 draws a JPEG again and saves it anew, so its pixels change and it loses a little quality; ExifTool and hexscope copy the picture's bytes as they are.
• HEIC, an iPhone's photos: MAT2 does not clean them. ExifTool and hexscope do.
• WebP: MAT2 failed on all three WebP files.
• TIFF: ExifTool left the camera's make and model, the program and the date in the first image directory.

3 Videos

Every MP4 and MOV keeps a creation and a change time in its own headers. ExifTool can change them but not remove them, so they stay in its copies; hexscope sets them to zero. MAT2 cleaned the MP4 files but refused the MOV, an iPhone's video, and the M4A recording.

4 PDFs

• ExifTool adds its changes after the original, which stays in the file: hexscope found the original titles, dates and programs, and the earlier versions, in every copy. ExifTool says so itself when it writes one: its PDF edits are reversible, and deleted tags may be recovered.
• MAT2 turns each page into a picture. Nothing identifying stays, and nothing can be selected, searched or read aloud either.
• hexscope writes the PDF anew with only what its pages use, and takes out text hidden under black boxes, which no metadata tool looks for.

5 Word, Excel, PowerPoint and LibreOffice

ExifTool does not write these files. MAT2 cleaned some, but left the comments in LibreOffice files, the picture of the first page in a Word document saved on Windows, and the folder Excel saved a workbook in, which holds the Windows account's name. It refused three files for a part it did not expect, which is its careful choice when it is not sure. hexscope cleaned all of them; it keeps, and says it keeps, what is part of the content: hidden sheets and speaker notes.

6 What it found in hexscope

ExifTool read three things in an old Word file (.doc) that hexscope did not show or remove: the dates and the revision number Word keeps in the document's own record, beside its properties. hexscope now reads them, with the list of who saved the document and where, and its clean copy blanks them.

7 Which to use

• ExifTool reads more formats and tags than any other tool, a camera's raw files among them. Use it to see everything a photo says, and in scripts.
• MAT2 fits a Linux desktop, and refuses a file rather than half-clean it. Expect JPEGs to be re-encoded and PDFs to lose their text.
• hexscope runs in a browser with nothing to install, shows what a file says before it removes it, cleans PDFs and documents without redrawing them, and finds text under black boxes. Its command line tool and GitHub Action do the same in scripts and CI.

Whatever you use, check the copy before you send it. No tool removes what the picture or the page shows.

Check a file now Your files never leave your browser.