DOCX to PDF conversion causes incorrect equation formatting and corrupted Visio diagrams

Start by tracing Filee’s DOCX-to-PDF conversion entry point and how it invokes the LibreOffice backend; the issue does not name source files or tests. Use the attached text.docx and compare the generated PDF with text.pdf, focusing on equations on page 1 and the Visio diagram on page 3. Done means equation and diagram formatting are preserved more faithfully, with the pagination difference investigated.

What happened?

When converting a Microsoft Word document (.docx) to PDF using Filee (LibreOffice backend), I encountered two issues related to mathematical equations and embedded Visio diagrams.

The same Word document can be exported to PDF correctly using Adobe Acrobat PDFMaker, but the PDF generated by Filee does not fully preserve the original formatting.

Issue 1: Mathematical equations are not rendered correctly

The Word document contains numerous mathematical equations created using Microsoft Word's built-in equation editor, including fractions, subscripts, superscripts, Greek letters, matrices, and multi-line equations.

When converting the document using Filee:

• Mathematical equations have noticeable differences in font size, spacing, and symbol positioning compared with the original Word document and the Acrobat PDFMaker output.
• Some subscripts, superscripts, and mathematical symbols appear improperly sized or positioned.
• The overall equation layout is not faithfully preserved, affecting the readability and visual consistency of technical documents.
• The document pagination also changes. The Acrobat PDFMaker output contains 6 pages, while the Filee output contains 7 pages.

Issue 2: Embedded Microsoft Visio diagrams become corrupted

The Word document also contains diagrams created in Microsoft Visio and directly copied and pasted into Word, including electrical engineering control block diagrams.

When exporting the document through Filee:

• Text labels and mathematical symbols inside Visio diagrams become distorted, incorrectly scaled, or overlapped.
• Some characters are rendered incorrectly. For example, the "PI" labels inside the control blocks appear as distorted or incorrect characters.
• Subscripts and variable names inside the diagrams overlap with other graphical elements.
• The connections and basic graphical structure are largely preserved, but the embedded text and symbols are visibly corrupted.

This issue is particularly noticeable in the voltage and current control block diagram on page 3 of both PDF outputs.

The same diagram is rendered correctly in the Adobe Acrobat PDFMaker output.

What did you expect?

I expected the PDF generated by Filee to preserve the original Word document as closely as possible, particularly:

1. Mathematical equations should retain their original formatting, including font sizes, Greek letters, fractions, subscripts, superscripts, and equation alignment.
2. Embedded Visio diagrams should retain their original appearance, including text labels, mathematical symbols, relative positioning, and graphical elements.
3. The document layout and pagination should remain consistent with the original Word document, subject to normal differences between rendering engines.

Ideally, the PDF generated by Filee should have visual quality comparable to the output generated by Adobe Acrobat PDFMaker.

Steps to reproduce

1. Open the attached text.docx in Microsoft Word.
2. Observe the mathematical equations and embedded Visio diagrams in the original document.
3. Export the document to PDF using Adobe Acrobat PDFMaker. The reference output is text.pdf.
4. Convert the same text.docx to PDF using Filee with the LibreOffice backend. The resulting output is text1.pdf.
5. Compare the two generated PDFs:
   • On page 1, compare the formatting and spacing of the mathematical equations.
   • On page 3, compare the voltage and current control block diagram, especially the embedded text and "PI" labels.
   • Compare the equation layouts and overall pagination throughout the document.

Files involved

Three files are provided for comparison:

• text.docx —
• text.docx
• text.pdf —
• text.pdf
• text1.pdf —
• text1.pdf
  These files contain identical source content and can be used to reproduce and investigate the issues.

Environment

• Filee version: 2.0.0
• Operating system: Windows 11, version 26H2
• Input format: Microsoft Word DOCX.
• Reference PDF converter: Adobe Acrobat PDFMaker 26 for Word.
• Filee conversion backend: LibreOffice.
• LibreOffice version reported in the generated PDF metadata: 26.2.6.3 (X86_64).

Additional information

The original DOCX contains Office Math equations as well as an embedded Microsoft Visio drawing object with an EMF graphical representation.

The problems may be related to how LibreOffice imports and renders Word equations and embedded Visio/EMF content. However, I have not determined whether the root cause is LibreOffice itself, font substitution, or Filee's conversion pipeline.

This issue significantly affects the usability of Filee for academic papers, engineering reports, and other technical documents containing mathematical equations and vector diagrams.

It would be helpful if Filee could improve conversion fidelity for these elements or provide an alternative PDF conversion method that better preserves Microsoft Word formatting.