Share Safe

Before you hand a file over, see what is hiding in it — listed per file with severity — and get a safe copy with everything removable removed. The copy is re-checked afterwards, and whatever is left comes with manual steps. Files are never uploaded.

On-device・No upload

Your files and input are processed only inside your browser on this device and are never uploaded to our servers.

Drop files here (multiple allowed)

or

Files are inspected on your device and never uploaded. The format is detected from the content.Accepts: Images, PDF, docx / xlsx / pptx, text / CSV / JSON / code, ZIPUp to 100 MB per filePaste (Ctrl+V) also works

Sample: a ZIP with a GPS JPEG, a docx with author and comments and a text file with API keys, an xlsx with a hidden sheet, and a PDF whose black box hides live text

Up to 50 files at a time, 100 MB per file. ZIPs are checked one level deep, up to 200 entries and 300 MB unpacked.

Safe copy settings

How to use

  1. Drop files or click “Choose files” (images, PDFs, docx / xlsx / pptx, text and ZIPs, mixed and several at once). Inspection starts at once and the format is detected from the content, not the extension.

  2. Read the overall verdict (Looks OK to share / Needs review / Dangerous) and the per-file risk list. Severity is High, Medium or Low; anything that could not be checked is marked “Cannot confirm”.

  3. Adjust the options if needed (masking personal information in text, removing PDF comments) and press “Make safe copies”. Only what can be removed automatically is processed, and each copy is re-checked right away.

  4. Look at what was resolved and what remains. Items that need manual work, such as tracked changes or hidden sheets, come with steps.

  5. Download copies one by one or as a ZIP, and save the risk report (CSV / JSON). The report holds file names, paths and the kind and count of each risk, never the values found (names, keys, numbers).

What it does

Looks in the right places per format

Images: GPS, date taken, serial numbers, editing software, embedded thumbnails. PDF: author, JavaScript, attachments, comments, form values, text under black boxes. Office: author, company, template path, custom properties, custom XML, comments (with author names), tracked changes, hidden sheets, rows/columns and defined names, pivot caches, external links, embedded objects, speaker notes, macros. Text: secrets and personal information.

Finds text under black boxes

It reads each PDF page’s drawing commands and looks for opaque dark rectangles drawn after the text that cover most of a character, so a box over one word of a long line is found. Boxes made from images, light-coloured, rotated or curved shapes and PDFs over 100 pages cannot be judged. Clipping is handled only by its bounding box.

Safe copies for what can be removed

Images and PDFs use the same removal as the metadata cleaners; Office files get author, last editor and company emptied and docx comments removed; text gets secrets, e-mail, phone, My Number–style and card numbers (Luhn check) masked; ZIPs are rebuilt without junk and with each entry made safe.

Re-checked after the copy

Each safe copy is inspected again and resolved and remaining items are shown side by side. Things that are not automated, such as accepting tracked changes, remain and come with manual steps.

The verdict is labelled as a guide

The verdict reflects only what could be detected. Information inside pages or images, custom formats and encrypted content are not checked. What could not be checked is “Cannot confirm”, never “safe”, and keeps the verdict at “Needs review”.

Use cases

  • Check a quote or report (docx / xlsx) for author names, comments, tracked changes and hidden sheets before sending it to a client
  • Remove location, shooting time and camera serial numbers from photos before posting them
  • Make sure a PDF you blacked out has no live text underneath before publishing it
  • Look for API keys, passwords, e-mail addresses and computer user names in code or logs before sending a ZIP
  • Drop __MACOSX, .DS_Store and similar junk from a folder before zipping it

Supported formats

Inspect and make safe copies

JPEG / PNG / WebP (metadata removal)PDF (document info, XMP, JavaScript, attachments, comments removed)docx / xlsx / pptx (author, company, template path and custom properties cleared; docx comments removed)Text, CSV, JSON, code (masking)ZIP (one level; junk removed, entries made safe)

Inspect only

HEIC / AVIF / TIFF (location and other metadata)doc / xls / ppt (legacy; format only, contents are not checked)Video, audio, 7z, RAR and others (format only)

Capability matrix

FeatureSupportNotes
Detect and remove image location, dates, serials and thumbnailsSupportedJPEG / PNG / WebP. HEIC, TIFF and others are detection only
Detect and remove PDF document info, JavaScript, attachments and commentsSupported—
Detect text left under PDF black boxesPartialOpaque dark rectangles only (several per path are fine). Image-based boxes, light-coloured, rotated or curved shapes and PDFs over 100 pages cannot be confirmed
Empty Office author / company / template path / custom properties; remove docx commentsSupported—
Detect custom XML, hidden defined names and comment author namesPartialDetection only. Remove them in Office (deleting the comments also removes the author names)
Detect tracked changes, hidden sheets, macros, pivots and notesSupportedRemoval is manual (accept or delete them in Office)
Remove xlsx / pptx commentsNot supportedDetection only. Delete them in Office
Detect and mask secrets and personal information in textSupportedHeuristic. Misses and false positives happen
Inspect and rebuild ZIPsPartialOne level. Nested ZIPs and encrypted files cannot be confirmed
Detect faces or text shown inside images or PDFs (OCR)Not supported—
Virus scanningNot supported—

Limits & notes

  • Up to 50 files at a time, 100 MB per file, processed one at a time.
  • ZIPs are inspected one level deep (200 entries, 50 MB per entry, 300 MB unpacked in total). Anything beyond that, and nested ZIPs, is “Cannot confirm”.
  • Text is inspected up to 5 MB; larger files get no safe copy.
  • The black-box check covers the first 100 pages of a PDF. Image-based boxes, light-coloured rectangles, rotated or curved shapes are not judged.
  • The verdict is a guide and cannot find every risk. Information inside pages or images, custom formats and encrypted content are not checked.
  • Accepting tracked changes, deleting hidden sheets, removing macros, deleting xlsx / pptx comments and applying redactions are not automated.

Privacy & safety

Reading, inspecting, rewriting and zipping all happen in your browser. Files, their contents and names are never sent to or stored on a server, and the values found are never recorded in analytics.

FAQ

What should I check before sending a file?
Photos: location and date taken. PDFs: author, comments and text under black boxes. Word / Excel: author, comments, tracked changes and hidden sheets. Text and ZIPs: API keys, e-mail addresses and computer user names. This tool checks all of these and lists them by severity.
Can text still be read under a PDF black box?
Yes. If a black rectangle was just drawn over the text, the text stays in the file and can be copied or searched. This tool detects dark rectangles drawn after the text that cover it. Image-based boxes cannot be judged.
Does “Looks OK to share” mean it is safe?
No. It means no risk was found in what could be checked; it is not a guarantee. Information inside pages or images is not checked, so look at the files yourself too.
Can it remove Word tracked changes and comments?
Comments in a docx are removed in the safe copy. Tracked changes need your decision on what to accept, so they are not processed automatically; you are told to accept them under Review in Word.
Does it look inside ZIPs?
One level deep: images, PDFs, Office files and text are each inspected, and a ZIP is rebuilt without junk files and with the contents made safe. Nested ZIPs and encrypted files cannot be confirmed.
Are my files uploaded?
No. Everything is processed in your browser and nothing is sent to a server.

Last updated: