Download Latest Version openpublisher5g_amd64-windows.exe (81.1 MB)
Home / 5.11.14
Name Modified Size InfoDownloads / Week
Parent folder
OpenPublisher5-i386-windowsXP.exe 2026-10-24 56.5 MB
openpublisher5_amd64-windows-portable.exe 2026-10-24 183.2 MB
openpublisher5_amd64-windows.exe 2026-10-24 165.2 MB
openpublisher5_amd64-macos.dmg 2026-10-24 233.1 MB
openpublisher5_amd64-linux.run 2026-10-24 193.4 MB
openpublisher5_amd64-linux.rpm 2026-10-24 215.4 MB
openpublisher5_amd64-linux.flatpak 2026-10-24 153.6 MB
openpublisher5_amd64-linux.deb 2026-10-24 156.2 MB
openpublisher5_amd64-linux.AppImage 2026-10-24 166.4 MB
openpublisher5_amd64-freebsd.pkg 2026-10-24 185.5 MB
5.11.14_ Dual-Canvas _.pub_ _ _.docx_ Graphics Preservation, Differential Ink Sampling _ Multi-Column Table Fix source code.tar.gz 2026-10-24 23.4 MB
5.11.14_ Dual-Canvas _.pub_ _ _.docx_ Graphics Preservation, Differential Ink Sampling _ Multi-Column Table Fix source code.zip 2026-10-24 23.7 MB
README.md 2026-10-24 9.2 kB
Totals: 13 Items   1.8 GB 0

πŸš€ Summary

Version 5.11.14 resolves a critical regression in v5.11.13 where image-based Microsoft Publisher (.pub) publications, greeting cards, photo collages, and vector-styled Word (.doc / .docx) documents rendered blank or lost their background graphics and table structures when imported.

  1. Dual-Canvas Text-Free Background Engine (bgCanvas):
  2. Upgraded window.OpenPubImporter.processPdfPage in io/file-io.js to render a second, text-suppressed background canvas (bgCtx.fillText = () => {}; bgCtx.strokeText = () => {};) alongside the full analysis canvas.
  3. Preserves 100% of raster photos, collages, greeting card artwork, vector backgrounds, gradients, borders, and table grids at zIndex: 1 while preventing PDF text from being drawn onto the background layer in the first place.
  4. Eliminates both blank pages on image-heavy .pub publications and double-vision ghost text on hybrid .pub, .doc, and .docx documents, without relying on optional server-side ODG extraction packages (adm-zip / @xmldom/xmldom).
  5. Added a try / catch fallback in processPdfPage that always returns the rendered high-resolution page canvas at zIndex: 1 if an unexpected exception occurs during text extraction, guaranteeing an imported page can never render blank.
  6. Differential Core-Glyph Ink Color Sampling:
  7. Replaced the single-scanline < 240 dark-pixel filter in processPdfPage with differential pixel comparison between fullImgData (rendered with text) and noTextImgData (rendered without text) across each text item's bounding box.
  8. Identifies the maximum contrast delta (maxDiff) between the glyph and its exact underlying background pixel and averages the unblended core stroke pixels (diff >= Math.max(24, maxDiff * 0.65)).
  9. Accurately extracts pure white and light-cream text on dark or textured backgrounds as well as dark charcoal and colored text on white pages.
  10. Multi-Column Table & 2-Panel Card Layout Preservation:
  11. Updated reconstructPublisherElements in io/file-io.js so shortcut-sheet heuristics (right-side callout merging and 210px inline key gap bridging) only activate on pages that actually contain shortcut key items (hasShortcutKeys).
  12. On standard .pub, .doc, and .docx pages, horizontal span grouping is restricted to Math.max(seed.fontSize * 1.5, 20), preventing multi-column tables from collapsing into a single column and keeping multi-panel greeting cards accurately positioned on their respective panels.
  13. Targeted Inline Icon Inpainting (eraseRegionsFromCanvas):
  14. Added eraseRegionsFromCanvas in io/file-io.js to cleanly erase promoted inline raster Windows key icons from bgCanvas on shortcut instruction sheets (GitHub Issue [#10]), leaving pure white text pages with nonWhiteRatio === 0 free of any background image layer.
  15. Safe .opub Loader & Legacy Sanitizer Guards:
  16. Updated sanitizeLegacyOpubData in io/file-io.js so full-page background elements are only stripped on legacy .opub files that actually contain shortcut key sheets (hasLegacyShortcutKeys), preserving backgrounds on all other .opub publications.
  17. Guarded DashboardSystem.close() inside window.handlePublisherFileLoad (io/publisher-file-loader.js) with a try / catch block so .opub file loading never halts if the dashboard overlay DOM node is absent.
  18. ServerSide Optional ODG Guard:
  19. Updated ServerSide/server.js to skip the secondary --convert-to odg headless LibreOffice execution when optional adm-zip / @xmldom/xmldom packages are not installed, cutting conversion time in half on standard deployments.
  20. Version Synchronization:
  21. About dialog subtitle in extensions/ui-interactions.js updated to Free Online Desktop Publishing Tool β€’ v5.11.14.
  22. Application headers in core/boot.js, core/core.js, extensions/print-engine.js, extensions/ui-interactions.js, and elements/wordart/legacy-wordart.js bumped to v5.11.14.
  23. Service Worker cache upgraded to open-publisher-v5-11-14 in sw.js.
  24. Asset cachebusters across all 85 resource tags in index.html updated to ?v=5.11.14.

πŸ” Deep Dive: Root Cause Analysis & Architectural Fixes

1. Why Image-Heavy .pub and Vector .docx Files Rendered Blank in v5.11.13

  • The Problem:
  • When LibreOffice converts image-based .pub publications (such as posters, photo collages, CD covers, and greeting cards) to PDF, the resulting PDF pages often contain 0 PDF text operators and 1 to 34 raster image XObjects (paintJpegXObject / paintImageXObject). Similarly, styled Word documents often draw colored page backgrounds and multi-column table cells using PDF vector path fill operators (0 image XObjects).
  • In v5.11.13, processPdfPage removed the canvas.toDataURL('image/jpeg', ...) background element (zIndex: 1) and relied on options.structuredPage.images returned by the server's optional ODG parser. Whenever adm-zip and @xmldom/xmldom were not installed on the server (or when importing via /api/convert-doc), structuredPages was null, causing image-only .pub pages to emit 0 elements (blank white pages) and stripping vector backgrounds and photos from hybrid pages.
  • The Solution:
  • processPdfPage now renders a text-free bgCanvas by intercepting bgCtx.fillText and bgCtx.strokeText during PDF.js rendering.
  • All raster images, vector fills, gradients, borders, and table cells are rendered onto bgCanvas with 100% fidelity and zero baked-in PDF text.
  • Whenever textElements.length === 0 || nonWhiteRatio > 0.0006 || (pageHasImages && !hasShortcutKeys), bgCanvas.toDataURL('image/jpeg', 0.90) is emitted at zIndex: 1 behind the editable HTML text elements.

2. Differential Core-Glyph Ink Sampling vs. Fixed < 240 Thresholding

  • The Problem: Sampling a single scanline for pixels darker than RGB(240, 240, 240) failed on light/white text over dark backgrounds, turning white text dark or washing it out by averaging anti-aliased edge pixels with the dark background.
  • The Solution: Because processPdfPage has both fullImgData (with text) and noTextImgData (without text), every pixel where fullImgData differs from noTextImgData inside a text item's bounding box is guaranteed to belong to a glyph stroke. By averaging only pixels with diff >= Math.max(24, maxDiff * 0.65), processPdfPage captures the unblended center of the glyph stroke for crisp white, cream, or dark ink reproduction.

πŸ“ Files Modified in 5.11.14

File Changes
io/file-io.js Added dual-canvas text-free background rendering (bgCanvas), differential core-glyph ink color sampling, eraseRegionsFromCanvas(), multi-column table gap preservation in reconstructPublisherElements(), shortcut-key guard in sanitizeLegacyOpubData(), and try / catch canvas fallback in processPdfPage().
io/publisher-file-loader.js Wrapped DashboardSystem.close() in a try / catch guard inside window.handlePublisherFileLoad.
ServerSide/server.js Guarded optional --convert-to odg LibreOffice execution with if (AdmZip && DOMParser) so standard deployments do not run a redundant conversion pass.
extensions/ui-interactions.js Updated About dialog subtitle to Free Online Desktop Publishing Tool β€’ v5.11.14. Bumped header and console log to v5.11.14.
core/boot.js Bumped Application Bootstrapper header to v5.11.14.
core/core.js Bumped Core Engine header to v5.11.14.
extensions/print-engine.js Bumped Print Engine headers to v5.11.14.
elements/wordart/legacy-wordart.js Bumped WordArt Engine headers to v5.11.14.
sw.js Bumped Service Worker cache name to open-publisher-v5-11-14 and header to v5.11.14.
index.html Updated all 85 asset cachebusters from ?v=5.11.13 to ?v=5.11.14.

βœ… Verification Checklist

  • [x] Image-Based .pub Publications Verified: Full-page posters, multi-page greeting cards, multi-photo collages, and CD covers render with 100% intact graphics.
  • [x] Hybrid .pub Greeting Cards Verified: Two-panel floral and textured cards preserve their full background artwork at zIndex: 1 with editable light-colored text lines on the foreground layer and zero ghost text.
  • [x] Vector-Styled .docx & .doc Documents Verified: Dark-background and multi-column table Word documents retain their full vector backgrounds, table cell grids, and crisp white/colored editable text elements with zero ghost text.
  • [x] GitHub Issue [#10] Shortcut Flyers Verified: All editable text elements, inline Windows key icons, centered lines, and right-aligned callouts render cleanly on a pure white page without background ghost text.
  • [x] Version & Cachebusters Synchronized: index.html, sw.js, core/boot.js, core/core.js, extensions/ui-interactions.js, extensions/print-engine.js, and elements/wordart/legacy-wordart.js are all synchronized to v5.11.14.

πŸŽ‰ Version 5.11.14 - Open Publisher

Source: README.md, updated 2026-10-24