Stop fixing LaTeX one by one.
Drop your file, get every equation.
Pasted a ChatGPT answer, built a worksheet, or wrote a paper in Word? Convert every piece of LaTeX in your document at once — no equation editor, nothing leaves your browser.
⚠ Your document must already have math in LaTeX delimiters — $…$, $$…$$, \(…\), \[…\], or [ … ]. Bare LaTeX like \frac{1}{2} is left as plain text.
Not in that format? Ask ChatGPT: Write every formula in LaTeX — $...$ for inline, $$...$$ for display. Never write math as plain text.
Your file is read in the browser, never uploaded to us. Why →
[\...] too
What LaTeXFlow Scan does
When you write math as LaTeX inside a Word or Google Docs file — for example $\nabla \cdot \vec{F}$ or $$E = mc^2$$ — it stays as plain text. It looks fine while you type, but the moment you build slides, hand the file to a co-author, or export a PDF, every formula shows up as raw backslashes and braces.
LaTeXFlow Scan reads a .docx file, finds every piece of LaTeX in it, and gives you back the same document with each formula replaced by a rendered image. No equation editor, no retyping, and no formula ever leaves your browser — the whole conversion runs on your own machine.
How it works
- Bring your file in. Drop a
.docxonto the page, or use Import from Google Drive to pull a Google Doc in directly. The Drive import uses the limiteddrive.filescope, so the tool can only see the single file you pick — never the rest of your Drive. - Detection. The scanner walks the document and collects every span wrapped in a LaTeX delimiter (see the table below).
- Review. Each detected formula gets a live preview. Skip anything that was caught by mistake (a stray
$100, say), and resolve ambiguity cards where a paragraph has an unclosed$— the tool asks you to pick the boundary instead of guessing. - Render & export. Press Render PNG · Export and download a
.docxin which every formula is now a crisp PNG. The original LaTeX is preserved in each image's alt text, so the document can be converted back later.
Supported LaTeX delimiters
A formula is only detected if it already sits inside one of these delimiters. Bare LaTeX with no delimiter — \frac{1}{2} on its own — is treated as ordinary text and left untouched.
| Delimiter | Example | Detected |
|---|---|---|
$ … $ | $x^2 + y^2$ | Inline math |
$$ … $$ | $$E = mc^2$$ | Display math |
\( … \) | \(a_i\) | Inline math |
\[ … \] | \[\int_a^b f\,dx\] | Display math |
[ … ] | [\lim_{x\to0}\frac{\sin x}{x}] | Bracket mode — for AI-chatbot output whose backslashes were stripped by markdown. On by default; toggle it off if it over-detects. |
When people use it
Google Docs class notes → clean slides
Students jot equations as LaTeX while taking notes, then need presentable formulas for a presentation. Import from Drive, convert, done — no re-typing before the deadline.
A Word paper full of equations
A draft written in Word with dozens of inline formulas would take an afternoon to fix by hand. One pass converts them all. (If you need true equation numbering and cross-references, write the paper in LaTeX instead — this tool is for turning existing documents into clean, shareable output.)
AI chatbot answers pasted into a doc
Paste a ChatGPT or Claude answer into a .docx and the display math often arrives as [ … ] — the outer backslashes were eaten by the chat window's markdown. Bracket mode catches exactly that case. Tip: ask the chatbot for $…$ and $$…$$, and paste it as text — don't ask it to “make a .docx”, or it exports Word math objects this tool can't read.
Frequently asked questions
AIMATH_FORMULA::v1:: tag), so a formula can be recovered — for instance with the LaTeXFlow Google Docs Add-on.$$…$$) sits within a single paragraph — a formula split across paragraphs isn't detected. Native Word/Docs equation objects aren't read either; the tool works on LaTeX written as text.$100 got picked up as math.$ looks like an opening delimiter. Just Skip it in the Review step, or resolve it from the ambiguity card..docx needs no account at all.Privacy
The document you convert never leaves your browser. To train future equation-recognition models, the LaTeX source and rendered image of each equation are collected anonymously — never your document text, file name, or email. Google Drive import is scoped to drive.file — the single file you choose in the picker, nothing more. Full details, and how to opt out, are in the Privacy Policy.