A Word document that lost its images but kept the text

The document opens, every paragraph is where you left it, and each picture is now an empty box, a red X, or a caption with nothing above it. The good news is that the text surviving tells you a lot: in most cases the images are still inside the file and only the references to them are broken.

A DOCX is a zip archive, and the pictures live in their own folder inside it. You can look in that folder yourself in about a minute, without Word and without any repair software.

Why the text survives and the pictures do not

A .docx file is a zip containing several parts: the text as XML in word/document.xml, the styles, the relationship map in word/_rels/document.xml.rels, and every embedded picture as an ordinary file in word/media/.

The text is one continuous stream, so a small amount of damage tends to either kill the whole document or leave the text intact. The images are separate entries, each one pointed at by an ID in the relationship map. Break the map, or lose one entry from the zip directory, and Word still renders the text perfectly while finding nothing at the other end of each picture reference. That is the empty box.

Three common causes:

• An interrupted save or a crash mid-write, which can leave the zip directory out of step with the actual entries.
• A copy that did not finish, particularly to or from a network share or an external drive. A file that copies to an external drive and arrives broken covers that pattern.
• The images were never embedded. If they were inserted with the link option rather than the insert option, the document only ever held a path to a file on the original machine.

That last one is worth ruling out early, because nothing can recover a picture that was never stored.

Look inside the file yourself

Work on a duplicate, always. Select the file in the Finder, press Command-D, and use the copy for everything below.

1. Rename the duplicate so it ends in .zip instead of .docx. The Finder will ask whether you are sure.
2. Double-click it. The Archive Utility expands it into a folder.
3. Open the folder that appears and look for word/media/.

If media is there and full of image1.png, image2.jpeg and so on, your pictures survived and this is a repair job rather than a recovery job. If the folder is missing or empty, the images were either linked rather than embedded, or genuinely lost.

If the Archive Utility refuses to expand it at all, the zip structure itself is damaged, and the document is in the territory covered by fixing a DOCX that Word says is corrupt.

The Terminal version, if you would rather not rename anything:

unzip -l ~/Desktop/report.docx | grep media

That lists the media entries with their sizes without extracting anything. Sizes of zero are a bad sign; real photos should be hundreds of kilobytes.

Get the pictures back out

Once you can see word/media/, you have every embedded image at full resolution, in whatever format Word stored it. They are named in document order, which makes reassembly less painful than it sounds.

For a document with a handful of images, copying them out and reinserting them by hand is often faster than any repair attempt, and the result is a clean file you can trust. Note that Word sometimes stores a picture in a format the Finder previews poorly, such as EMF or WMF from a pasted diagram, or TIFF from a scanner. Those convert cleanly to PNG when you need to reuse them elsewhere.

Repairing the document instead

If the document is long, or the layout matters more than the individual pictures, repairing it in place is the better path. A repair pass rebuilds the archive structure and the relationship map from what is actually present, so pictures that exist get reconnected and pictures that do not are dropped rather than left as broken boxes.

Manatee does that on the Mac in Repair mode: drop the document on it and it works from a duplicate, never touching your original, and reports the outcome plainly as Fixed, Partial or Not fixable. Partial is the common and honest answer here, for example a document that comes back with the text and eleven of fourteen images. What that verdict actually means is unpacked in partial recovery explained.

Word’s own Open and Repair, from the Open dialog, is worth trying first and costs nothing. It is better at structural problems in the XML than at missing zip entries, so the two approaches fail in different places and it is reasonable to try both.

When the images really are gone

Some cases are not recoverable, and it is better to know quickly:

• Linked rather than embedded images. Find the original picture files and reinsert them, choosing Insert rather than Insert and Link.
• A file that was truncated in transit. If the document is noticeably smaller than you remember, the tail of the archive is missing, and the media entries stored there went with it.
• A document that was recovered from a deleted state by any tool. What comes back is whatever the disk still held, and the gaps are permanent.

Manatee will not help with any of those, and neither will anything else. Data that is gone is gone, which is the uncomfortable half of why files get corrupted and what can actually be recovered.

Keeping it from happening again

A few habits reduce the odds considerably.

Save documents with images to local storage first and copy them to a share or a cloud folder when you are done, rather than editing in place across a network. Most broken-media reports trace back to an editing session over an unreliable connection.

Turn off the option to compress pictures on save if your workflow uses high resolution photographs, since that rewrite is another moment where a crash can leave the archive inconsistent.

Keep the source images in a folder next to the document while it is in progress. Reinserting from originals takes minutes; reconstructing from a damaged file can take an afternoon, and the reinserted version is better quality anyway.

Finally, treat any document that has behaved strangely once as suspect. Copy the text into a fresh document and rebuild rather than waiting for the second failure.

Questions

Can I recover the images if the document will not open at all? Often yes. The zip rename trick works even when Word refuses the file, because the Archive Utility does not care about the document XML. If the archive expands, look in word/media/ regardless of what Word says.

Why are the images larger inside the file than the ones I inserted? Word stores an embedded copy, and depending on the insert path it may store the full original rather than the on-screen version. That is why a document with a few photos can run to tens of megabytes.

Does the same approach work for PowerPoint and Excel? Yes. PPTX and XLSX are also zip archives, with media in ppt/media/ and xl/media/ respectively. The rename-to-zip inspection is identical.

Will reinserting the extracted images change the quality? No, as long as you insert the file from word/media/ directly and do not let a conversion step in between. Those files are the exact bytes the document was carrying.