wordpaste → OMML to LaTeX

Convert Word's OMML equations to LaTeX

Microsoft Word does not store equations as pictures or as LaTeX. It stores them as OMML — Office Math Markup Language — both inside a .docx (in word/document.xml) and on the clipboard when you copy. This page converts that OMML to LaTeX, in your browser, with no upload and no server.

MIT zero dependencies 3.6 kB gzipped runs in the browser

Try it

Paste an <m:oMath> fragment — straight out of a .docx, or from the HTML Word puts on the clipboard.

OMML in

LaTeX out

LaTeX appears here.

Rendered (KaTeX)

Nothing to render yet.

Using it in code

npm i wordpaste

import { ommlToLatex } from 'wordpaste';

const latex = ommlToLatex('<m:oMath>…</m:oMath>');
// '\\frac{a}{b}'

One function, string in and string out. It returns an empty string when the fragment cannot be parsed, so it never throws at you mid-paste.

If what you actually have is a paste from Word rather than a bare OMML fragment, use transformPastedHTML instead — it cleans the whole clipboard payload and converts the equations as part of that.

What it converts

OMML	Meaning	LaTeX
`m:f`	fraction — bar, `noBar`, `lin`, `skw`	`\frac`, `\binom`, `a/b`
`m:rad`	root, with or without a degree	`\sqrt`, `\sqrt[n]`
`m:sSub` `m:sSup` `m:sSubSup`	sub- and superscripts	`_{}` `^{}`
`m:sPre`	pre-sub/superscript, as in an isotope	`{}_{6}^{14}{C}`
`m:nary`	∑ ∫ ∏ with limits	`\sum_{}^{}`
`m:d`	delimiters, including ‖ ⌊ ⌈ ⟨ and empty sides	`\left\lfloor … \right.`
`m:acc`	accents — the character decides which	`\vec` `\bar` `\dot` `\hat` …
`m:bar`	rule above or below	`\overline` `\underline`
`m:m`	matrix	`\begin{matrix}`
`m:eqArr`	multi-line equation array	`\begin{aligned}`
`m:func` `m:limLow` `m:limUpp`	named functions and limits	`\lim_{}`
`m:nor`	a run marked upright — units, words	`\text{}`

Unicode maths glyphs Word emits as plain runs (× ÷ ≥ ∞ ∆ π θ …) are mapped to their LaTeX commands, because KaTeX will not read several of them raw.

What it does not do

It needs a browser. The parser is the native DOMParser. On the server, give it one from jsdom first — there is a snippet in the README.
It reads OMML, not MathML. LibreOffice and older Word paste MathML instead; transformPastedHTML handles that case, this function does not.
It is not a full ECMA-376 implementation. Rare constructs (m:groupChr, m:borderBox, m:phant) are descended into rather than translated, so their contents survive but their decoration does not.
It is not a sanitiser. Output is LaTeX text, but if you build HTML from it yourself, escape it.

Why this exists

Word puts an equation on the clipboard twice: once as a rasterised screenshot, and once as the real OMML. Almost every editor takes the screenshot, so the maths is never editable again. Reading the OMML instead is the whole point of wordpaste, and ommlToLatex is that part on its own, for when you are working with .docx files rather than a paste.