
LaTeX Word,
precisely.
Drop a .tex, .docx, or a .zip project — research papers, reports, theses, books, manuals, lecture notes, CVs, contracts, anything you write. LaTeX → Word is the high-fidelity direction: supported equations, tables, figures and BibTeX citations are preserved (upload the .bib alongside your .tex in a .zip to get a formatted References section). Word → LaTeX returns clean, compilable source — structure, math and tables come through, but expect some manual polishing of preamble, packages and fine layout. Every run ships a quality report listing anything the engine couldn't carry over. Used worldwide, with prices shown in your currency. See the compatibility matrix.
Conversion desk
.texWord.docx.docxLaTeX.texDrop your file here
or click to browse — direction is detected automatically
Also accepts .zip projects with figures & bibliography · Max 25 MB
Before / after
See what a real conversion looks like
A short LaTeX source with an equation and a table on the left; the exact rendered Word document on the right. No cherry-picking — this is the unedited output of the engine.
\documentclass{article}
\usepackage{amsmath,booktabs}
\title{Convergence of \(a_n\)}
\author{A. Ramanujan}
\begin{document}
\maketitle
\section{Result}
For any \(\epsilon > 0\), there exists \(N\) such that
\begin{equation}
\lim_{n \to \infty} \left(1 + \tfrac{1}{n}\right)^{n} = e.
\end{equation}
\begin{table}[h]
\centering
\begin{tabular}{lrr}
\toprule
\(n\) & \((1+1/n)^n\) & error \\
\midrule
10 & 2.5937 & 0.1245 \\
100 & 2.7048 & 0.0134 \\
1000 & 2.7169 & 0.0014 \\
\bottomrule
\end{tabular}
\end{table}
\end{document}Convergence of an
A. Ramanujan
1. Result
For any ε > 0, there exists N such that
| n | (1+1/n)n | error |
|---|---|---|
| 10 | 2.5937 | 0.1245 |
| 100 | 2.7048 | 0.0134 |
| 1000 | 2.7169 | 0.0014 |
Editable Word math (OMML), real Word tables, preserved section numbering.
Most things are free. Server-side work costs credits.
Free forever
LaTeX ↔ Word conversion in your browser, PDF ↔ Word, Excel → PDF, Markdown → Word, image merges. Unlimited, no credits used.
Costs credits
1 credit = 1 page (≈ 300 words) when you use server features: TikZ / diagram rendering, OCR, and quality-verified downloads. You always see the exact cost before you confirm.
Credits cost
₹50 per credit at base rate. Larger packs are discounted up to ~70%. Pay once — credits never expire.
New accounts start with a small free credit balance so you can try the paid features before buying.
More converters
Always free · no credits required · runs in your browser
Built for accuracy
Under the hood, textportdocx is Pandoc compiled to WebAssembly, wrapped in a custom pre- and post-processing pipeline — math normalization, macro expansion, table repair, a strict-retry pass that catches leaked LaTeX, and optional TikZ rendering for figures. The core conversion runs on your device; server features (TikZ, OCR, billing, account sync) run only when you opt in.
High fidelity
Preserves supported equations, tables, figures and cross-references. BibTeX citations are resolved into a formatted References section when you upload the .bib alongside the .tex in a .zip. Each run ships a quality report listing anything unsupported.
Private by default
Core LaTeX ↔ Word conversion runs locally in your browser. Optional server features (TikZ rendering, billing) only run when you enable them.
Instant results
After the first conversion, the engine stays warm — repeat conversions are nearly instant.
How it works
One clean pass, start to finish
- 01
Drop your file
A single .tex or .docx, or a .zip with your images and .bib. Nothing is uploaded for the core conversion.
- 02
The engine runs
Macro expansion, math normalization, table repair and a strict-retry pass catch anything that would leak raw LaTeX.
- 03
Review and download
You get the converted file plus a quality report listing every construct the engine could not carry over.
Questions
Everything people ask first
Equations become editable Word math (OMML), tables become real Word tables, and section numbering, cross-references and figures are carried over. Anything the engine cannot map is listed explicitly in the quality report after every run.