Metadata Cleaner

About Metadata Cleaner

Drop in photos (JPEG, PNG, HEIC, WebP, GIF, TIFF, RAW), video (MP4, MOV), audio (MP3, FLAC, WAV), PDFs, Word, Excel and PowerPoint files including legacy Office formats, SVG or ZIP. The tool first lists every piece of metadata it can read, item by item, and by default removes all of it. Uncheck only what you want to keep.

It works from a whitelist: only the structural data needed to render the file stays. Anything it cannot identify (an unknown APP segment, bytes tacked on after the end of a JPEG) is rated high risk and removed. The preview and the actual cleaning run the same code with different selections, so what you see in the list is exactly what happens to the file.

Cutting vs. overwriting: each format needs its own method

In JPEG, PNG, WebP, GIF and ZIP-based files, metadata sits in separate segments or chunks that can simply be cut out. The file gets smaller and the image data is untouched.

TIFF and RAW, HEIC, MP4 and MOV, and legacy Office (OLE) files are different. They contain absolute byte offsets, such as "pixel data starts at byte N". Cut something out of the middle and every later offset points to the wrong place. So here the tool keeps the length and overwrites the bytes with zeros. In MP4 and MOV, a metadata box is renamed to a "free" box of the same size (a standard box type that players ignore) and filled with zeros. These items are marked as scrubbed, and the file size does not change.

In HEIC, EXIF and XMP are stored as items, so only their data ranges are overwritten. The file's index still points to those ranges as EXIF and XMP items, and plain zeros make strict readers (Python's Pillow, for example) fail on a broken EXIF block, so the tool writes an empty EXIF with no entries and an empty XMP packet at the start and blanks the rest. HEIC handles rotation with a separate box rather than EXIF, so wiping EXIF does not turn the photo sideways.

What turns up in files that are not photos

Word, Excel and PowerPoint files are ZIP archives of XML. Beyond document properties (author, last saved by, company, total editing time), the tool finds the names and timestamps left by tracked changes and comments, revision save IDs (rsids) that can link documents edited on the same machine, local paths to templates and linked files (C:\Users\yourname\… reveals the account name), printer settings containing printer and computer names, and EXIF inside embedded images. Embedded images are cleaned recursively with the same rules.

For PDFs it covers the document info dictionary (title, author, producer), XMP, annotation authors and times, attached files and application leftovers. For video it covers user-data boxes holding device and location tags, plus the creation and modification times written inside the file. For audio, ID3 and similar tags and album art. File times and owner fields inside a ZIP are always reset when the archive is rewritten.

Microsoft's own Document Inspector covers similar ground from inside Office, and also finds hidden text and invisible objects. The two are complementary: the Inspector understands document content, this tool sees the raw package.

What stays by default, and why

Items under "Removing these changes how the file looks or behaves" are not personal but affect display. For JPEG, when the full EXIF block goes, the orientation value can be written back as a minimal EXIF segment holding only that tag. Removing an ICC profile can slightly desaturate a wide-gamut photo. Removing a GIF's loop setting can make an animation play once. The Adobe APP14 segment in JPEGs describes the CMYK color transform; removing it can invert colors, so the tool leaves it alone and does not list it.

What this tool cannot remove

Clearing metadata does nothing to what is visible. A view from a window, a street sign, a shipping label, a school uniform, a reflection in glasses or a window, or text on a monitor in the background can locate or identify someone. In documents, the tool strips reviewer names, but the tracked changes and comment text, hidden text and white-on-white text remain. Accept all changes and delete comments in the original application first.

Watch the file name too. An original name like IMG_20260922_211005.jpg carries the capture time, so the default save name is a neutral one that keeps only the extension, such as "image-cleaned.jpg". Turn on "Keep the original file name" and it saves as "cleaned_" plus the original, so check whether the date is still in it. Created and modified dates kept by the operating system live outside the file and are beyond this tool.

Files in an unrecognized format are marked as not cleanable and cannot be downloaded. Legacy Office and Hangul (OLE) documents can only have their summary information wiped; traces such as the last-saved-by list may remain, so the tool suggests re-saving as .docx and cleaning that. Encrypted PDFs must be unlocked first.

Worked examples

How to use it

1. Drop in one or more files (up to 200 MB each).
2. Review what was found and its risk level; uncheck anything to keep.
3. Check the save name, then download each file or all of them as a ZIP.

Frequently asked questions

Related guides

• What Your Photos Say About You: EXIF, XMP, IPTC and Location Privacy