LaTeX to DOCX maps our document engine's parsed document into Word XML, so math becomes OMML and LaTeX-only build behavior is discarded.
From real runs
| Element | Result | What happens |
|---|---|---|
| Title block | Kept | \title and \author were written into the DOCX body as a title paragraph and author paragraph. |
| Sections | Kept | \section became a Word Heading 1 paragraph and \subsection became a lower heading with bookmarks in document.xml. |
| Table of contents | Changed | The toc option created a Word TOC field marked dirty, so Word may need to update the field before page numbers show. |
| Equations | Kept | Inline and display math were encoded as Office Math XML rather than retained as literal LaTeX strings. |
| Tabular data | Kept | The three-column tabular environment became a Word table with a header row and right alignment in the numeric column. |
| Footnotes | Kept | \footnote generated a Word footnoteReference in document.xml and matching text in word/footnotes.xml. |
| Unresolved citations | Lost | The \cite{doe2026} command did not appear in the DOCX; the sentence kept two spaces where the citation had been. |
| Included graphics | Changed | A relative \includegraphics target became a Word hyperlink to assets/pixel.png, with no word/media file embedded and an image warning header. |
\title and \author were written into the DOCX body as a title paragraph and author paragraph.
\section became a Word Heading 1 paragraph and \subsection became a lower heading with bookmarks in document.xml.
The toc option created a Word TOC field marked dirty, so Word may need to update the field before page numbers show.
Inline and display math were encoded as Office Math XML rather than retained as literal LaTeX strings.
The three-column tabular environment became a Word table with a header row and right alignment in the numeric column.
\footnote generated a Word footnoteReference in document.xml and matching text in word/footnotes.xml.
The \cite{doe2026} command did not appear in the DOCX; the sentence kept two spaces where the citation had been.
A relative \includegraphics target became a Word hyperlink to assets/pixel.png, with no word/media file embedded and an image warning header.
We converted our LaTeX sample (217 B) with Document engine on 2026-09-11. It took 36 ms and produced a 11 KB Word DOCX file, 50× larger than the input. The engine raised no warnings.
Input · sample.tex · 217 B
9 non-empty lines, 22 words
Output · sample.docx · 11 KB
Word document: 4 paragraphs, 1 heading, 0 tables, 0 embedded images; fonts Consolas
pandoc
pandoc paper.tex -o paper.docx --tocReview citations, figure paths, and package-specific macros because pandoc is translating structure, not running a LaTeX build.
The HTTP API and the MCP server run the exact engine this page uses (Document engine); the response carries the same warnings the web converter shows. Check support first with GET /api/v1/pairs?from=latex&to=docx.
curl -X POST https://filetypeconverters.com/api/v1/convert \
-H "Authorization: Bearer $FTC_API_KEY" \
-F "file=@input.tex" -F "from=latex" -F "to=docx" \
--output output.docx/convert/latex-to-pdf after removing unsupported package and asset dependencies./convert/md-to-docx for cleaner Word styles.The tested equations were written as Office Math XML, which Word treats as equation content. They are not stored as the original LaTeX source.
No. In the unresolved citation test, the key did not appear in document.xml and the sentence contained blank spacing where the citation had been.
A relative image did not embed anything in word/media. The tested DOCX contained hyperlink text reading image and an external relationship targeting assets/pixel.png.
Our document engine creates a TOC field rather than calculating Word page numbers. Open the file in Word and update fields so the TOC reflects the document.