How to get the text and images out of a .pub file

You have a PDF of the 2013 newsletter and that is fine for reading. But now somebody wants the photograph from page two for the website, or the wording of the constitution that was printed in the back of the handbook, and a PDF turns out to be a poor place to keep either of those things.

That is the extraction problem, and it deserves its own answer.

Why a PDF is not enough on its own

A PDF records glyphs placed at coordinates on a page. It is an excellent record of what the page looked like and a mediocre container for what the page said.

Copy text out of one and you will typically find hyphenation from the original line breaks baked into words, two columns interleaved line by line, bullet characters arriving as stray symbols, and the occasional passage that comes out as nonsense because of how a font was embedded. It is all fixable with patience. It is not clean source text.

Images are worse. What you can get from a PDF is generally a re-compressed version at whatever resolution and quality the export used — not the original photograph that was placed into the document. If the original photograph was 3,000 pixels wide and the PDF was made for screen viewing, that detail has gone.

So: convert to PDF for the archive, absolutely. But extract the raw content as well, and do it before the original .pub becomes hard to open.

What is actually inside a .pub file

This is worth understanding, because it explains why some suggestions you will find online work and others cannot.

From Publisher 2.0 onwards, a .pub uses the Microsoft Compound File Binary Format — also called OLE2, or a compound document. Think of it as a tiny filesystem inside a single file: named streams of data arranged in a directory structure, rather like files in folders. Every such file begins with the same eight bytes, D0 CF 11 E0 A1 B1 1A E1. The same container was used by the older .doc and .xls formats.

The container itself is publicly documented. The part that matters is not. How Publisher encodes pages, the objects on them, their coordinates, and the relationships between them inside those streams is proprietary, and full documentation was never published. That gap is the whole reason reading a .pub is more work than reading a .docx.

Route 1: copy and paste out of Publisher (free, if you have it)

If Publisher still runs somewhere you can reach, this remains the most direct route: what you copy out comes from Publisher's own handling of the document, with nothing in between re-interpreting it.

• Text: click into a text box, press Ctrl+A to select the whole connected story rather than the visible frame, then copy and paste into Word. Publisher flows text between linked boxes, so selecting all from inside the story gets the continuation on later pages too — which a manual drag across one frame will miss.
• Images: right-click a picture and choose Save as Picture. This gives you the stored image rather than a screenshot of it. Do this for anything you might reuse; it takes seconds each.

Free, exact, and tedious at volume — the same trade-off as everywhere else in this topic.

Route 2: LibreOffice Draw (free, results vary)

Open the .pub in the free LibreOffice Draw and you can select text and copy it out, and right-click images to save them. Whether this works well depends on how completely the file imported in the first place, so check the result against the PDF or against your memory of the document rather than assuming everything came through.

It is free and it takes ten minutes to try, so try it before spending money on anything.

Route 3: a tool that parses the file directly (what we make)

This is the route for the situation the Publisher retirement is creating in bulk: you have the files and no Publisher, and you need the material out rather than a picture of the page.

Publisher Rescue reads the OLE2 structure itself, on your own computer, with no copy of Publisher involved. From your own .pub files it recovers:

• The text. Every text story in the document — the whole flow, not just what is visible in one frame — written out as a Word .docx, plus RTF and optionally a plain-text file.
• The images, as files you can use. PNG, JPEG and TIFF are written out exactly as they were stored. Windows bitmaps are rebuilt into openable BMPs. And WMF and EMF metafiles are decompressed and exported as real image files, rather than simply reported as present. Measured across the six genuine Publisher files this build has been run against: 20 image files were written — 15 PNGs, plus 5 BMPs rebuilt from bitmaps the documents used as pattern fills. Those particular files contained no metafile and no TIFF, so the metafile path has not met a genuine one; if vector logos are the reason you are buying, test your own file first.
• A PDF. The recovered content written into a PDF that ends with a notes page listing everything the tool could not rebuild. It is not a reproduction of your page design, and it says so on every page.
• The supporting detail. The page size read from the file; the number of pages actually rebuilt, which is the number of layout pages in that PDF rather than a claim about the original; and the Publisher format family the file was saved in — Publisher 2000, or Publisher 2002–2019, which share one version marker — rather than the exact release.
• A report of what happened. In plain language and in JSON, per file, stating what was found and what was not.

That WMF and EMF handling is worth dwelling on, because it is the part most often skipped. Metafiles are vector images and they were everywhere in the Publisher era — clip art, logos, diagrams, borders. Recovered properly, a vector logo scales to a banner without going soft. Recovered as a screenshot from a PDF, it does not. If your old documents contain a logo you still use, that alone can be the most valuable thing in the archive.

It runs on your own machine with no account — your files never leave it; the app goes online only to check your licence key, and works offline for up to 21 days between checks — which matters for the membership lists, pupil names and contact details that live in a surprising number of old newsletters. It is a one-time purchase. Saving the recovered files is what the £79 licence unlocks.

What to keep, in what form

For an archive you will still be able to use in ten years, keep all four of these side by side:

1. The original .pub. Never delete it. Formats get better support over time, not worse, and it is the only complete copy of the document that exists.
2. A PDF. The faithful record of how the page looked.
3. The text, as .docx or plain text. What the document actually said, in a form you can search and reuse.
4. The images, as individual files in a folder alongside. The ingredients, at their original quality.

Four small things instead of one. Disk is cheap; a photograph nobody kept a copy of is not replaceable.