1
Read the container
The importer opens the compound file and locates Publisher content, legacy Quill text and Escher drawing streams.
HomeNative PUB import evidence
Direct .pub migration, examined
Native PUB import is not PDF conversion or OCR
Plain-English answer: Megal's own importer reads supported structures inside the original Microsoft Publisher file, links their relationships and rebuilds them as separate Megal page objects. It does not first render the page to PDF or ask OCR to guess text from pixels.
That object-aware route matters when the goal is continued editing rather than a frozen copy. It also has limits: direct parsing does not make every proprietary Publisher feature reproducible, so important files still need page-by-page review.
Disclosure: These are Megal Software Limited's internal engineering results, not an independent certification or a guarantee for an arbitrary file. Evidence snapshot: 29 July 2026.
Output access: Free accounts can use a watermarked Print Preview. Actual printing and PDF creation, export or download require an active Professional subscription. Subscribe now for full unrestricted access.
How the importer works
From Publisher streams to an editable page model
A .pub file is a compound binary document containing several related data streams. Megal works through those records instead of treating the complete page as one picture.
2
Parse records
Dedicated parsers resolve supported pages, text stories, formatting, pictures, shapes and drawing properties.
3
Link relationships
The linker associates Publisher identifiers, stories, frames, pictures and drawing records where the evidence is sufficient.
4
Build page objects
The converter creates Megal pages and separate text-box, image, shape and table objects with supported geometry and styles.
Object-aware does not mean format-complete.
It means the importer attempts to preserve resolved document structure as objects. Ambiguous records may use bounded layout heuristics; unsupported content may be simplified, omitted or reported as a warning, so the result must still be reviewed rather than treated as exact.
Three different migration routes
Direct PUB parsing compared with PDF and OCR
These are workflow categories, not claims about any named competitor's undisclosed implementation. If another product does not document its engine, ask the vendor and test the resulting objects.
| Question | Direct, object-aware PUB import | PDF-based workflow | OCR from a page image |
|---|---|---|---|
| What is read? | The original .pub file and supported Publisher records. | A rendered or exported PDF, whose content depends on how that PDF was made. | Pixels in a scan, screenshot or rendered page image. |
| How is text recovered? | From supported text stories, runs, paragraphs and frame relationships. | From text and font data retained in the PDF, or by later reconstruction. | Characters and blocks are recognised statistically from their appearance. |
| What happens to objects? | Resolved text boxes, images, shapes and tables become separate output objects. | Object boundaries and reading order may need to be inferred from the rendered output. | Original Publisher object identities and relationships are not present in the pixels. |
| Main strength | Continued editing with more of the source document's structure available. | Preserving a stable finished-page appearance can be useful when editing is secondary. | Recovering visible text when only a scan or image survives. |
| Main limitation | Proprietary or unsupported Publisher features can still differ or be lost. | A visually faithful page is not automatically a faithful editable Publisher document. | Text, style, layers and geometry must be guessed; recognition errors are possible. |
Recorded engineering evidence
What the current test snapshot actually establishes
Megal uses two separate test layers: a large structural-import audit and a smaller rendered-page comparison. Their results answer different questions and should not be combined into one “accuracy” percentage.
9,866 .pub files
Every file in the recorded template corpus was passed through the native importer.
11,974 pages
The audit received non-empty documents with positive, finite page dimensions.
589,999 objects
The resulting page models contained separate shapes, text boxes, images and tables.
0 structural failures
No audited file threw an import exception or failed the audit's structural checks.
Numbers with their definitions attached
Corpus and visual-comparison results
Snapshot generated 29 July 2026. The structural audit covers 9,866 template files; the separate visual comparison covers 369 engineering comparison files. Both populations are held by Megal Software Limited, and the results have not been independently audited.
| Test layer | Scope | Recorded result | What it does not prove |
|---|---|---|---|
| Structural import audit | 9,866 .pub files; 11,974 output pages; 589,999 output objects; 95,795 text stories. | 0 failures. The object mix was 464,192 shapes, 95,802 text boxes, 29,995 images and 10 tables. There were 230 importer warnings across 221 files. | It does not prove visual fidelity, text reflow accuracy or support for every Publisher feature. |
| Publisher-render comparison | 369 .pub files and 442 pages with Microsoft Publisher PNG baselines and Megal-rendered counterparts. | 0 missing or failed comparison fixtures; 4.35% average pixel-mismatch ratio; 2.19% mean absolute colour difference; 88.24% average foreground intersection-over-union. | The comparison set is not claimed to be a random sample, and an average can hide a poor result on an individual page. |
How to read the structural pass
A structural pass means the importer completed, returned at least one page, produced positive finite page dimensions and gave each output object numeric, finite, non-reversed bounds. It is a useful guard against crashes and malformed output; it is deliberately not labelled “pixel-perfect.”
The 230 warnings were retained rather than hidden. Warnings can record parser decisions such as unresolved styled text or ignored Publisher scratch and pasteboard pages. A warning does not necessarily make the visible publication unusable, but it is a reason to review the affected file.
How to read the visual figures
Microsoft Publisher exported the reference pages as PNG images. Megal rendered the imported counterparts, the images were normalised to the same size and the comparison measured changed pixels, average channel difference and overlap of non-white foreground content.
The 4.35% mismatch figure uses a pixel-comparison threshold and excludes anti-aliasing differences. It should not be inverted and advertised as a “95.65% accuracy” score. The 88.24% foreground intersection-over-union is a different measure of overlapping visible content, not an editability score.
Why publish an imperfect number?
A measured mismatch is more useful than saying imports “look the same.” It shows the current engine is compared with Publisher-rendered baselines while acknowledging that visible differences remain.
Supported structure, not a flattened screenshot
What Megal attempts to reconstruct
The exact result depends on the records present in a file, but the import model has distinct handling for these parts of a publication.
Text and stories
Text boxes, styled runs, paragraphs, columns, linked-frame references, alignment, spacing and selected fitting information where resolved.
Geometry and drawing objects
Page-relative bounds, stacking order, rotation, flips, shapes, lines, fills, gradients and selected clipping or wrapping paths.
Pictures and tables
Embedded picture data, supported crop and mask information, plus table rows, columns, cells and basic cell styling when recognised.
The honest limitations
Direct parsing is a better starting point, not a print guarantee
- Publisher is proprietary. Some records, effects and relationships have no complete or safe mapping into a browser editor.
- Fonts affect layout. If the original font is unavailable or renders differently, line breaks, overset text and spacing can change.
- Complex content needs inspection. Legacy WordArt, unusual effects, grouped objects, tables, text wrap and master-page behaviour can require correction.
- Test averages are not your file. The visual metrics include both close matches and outliers. A representative real document is the best evaluation.
- Import is a migration snapshot. Save reviewed work as an editable .mpub document. Later changes to the original .pub file are not automatically merged into that project.
Keep the source .pub file, compare every important page with Publisher's rendering or a trusted printout, and proof the final export at 100% before distribution or print.
How to compare any PUB importer
- Ask whether the product parses the original .pub file or first creates a PDF or page image.
- After import, try selecting a paragraph, picture and vector shape separately.
- Check page count, dimensions, facing pages, object order and linked text frames.
- Replace text and images to see whether the result is genuinely reusable.
- Review warnings and unsupported features rather than accepting a preview alone.
- Test several representative files, including the most complex one you must keep using.
Common questions
Direct PUB import, PDF and OCR
Does Megal convert a PUB file to PDF before importing it?
No. Megal reads supported streams in the original .pub file and maps resolved content into its own page-object model. PDF export is not an intermediate import step.
Does Megal use OCR to read PUB files?
No. OCR recognises characters from pixels. Megal's PUB importer parses supported file records, text stories, object data and relationships from the original document.
Is a PDF workflow always worse?
No. PDF is often a good choice for preserving and sharing a finished page. It is a different objective from recovering Publisher structure for continued editing. Choose the route that matches the work you need to do next.
Does direct PUB parsing make every import exact?
No. Unavailable fonts, legacy effects, complex relationships or unsupported records can change a result. Direct parsing provides more source structure to work with, but every important page should still be proofed.
How can I verify another product's importer?
Look for a technical statement from its developer, ask whether PDF or OCR is used, and inspect the output. Being able to select and edit separate objects is stronger evidence than a page preview alone.
Test the objects that matter in your own file
Open a representative .pub file, inspect its text, pictures and shapes, then compare every page before choosing a migration route.