Both directions, in this tab. Your file never leaves it.
Every PDF-to-Word site on the internet wants your document on their server.
People put contracts, medical letters and payslips through them without
thinking about where the file went. This one has no endpoint to send anything
to — turn off your wifi after the page loads and it still works.
It tells you how well it will go, before you commit.
PDF was designed to look the same everywhere, not to be edited. Some PDFs
carry a hidden description of their own structure — which text is a
heading, what belongs in which table cell, what order it reads in. Anything
saved out of Word, Google Docs or LibreOffice usually does. For those, this is a
translation, and it is very good.
The rest — scans, LaTeX papers, output from older tools — are just
positioned letters on a page. Nothing in the file says where a paragraph starts,
so the structure has to be worked out from the layout. That is a reconstruction,
and no converter on earth does it perfectly. The difference here is that you are
told which of the two you have before you convert, and what to check
afterwards.
1. Open a file
Drop a PDF or a .docx here, or click to chooseNothing is uploaded — it is opened inside this tab
2. How well this will go
Headings not right? Nudge it, and this browser will
remember for the next document from the same source.
Preview
The first part of what came out, so you can see the structure before you open it.
The honest notes
Nothing is uploaded. Static page, no endpoint, no account. Both directions run in your browser.
PDF → Word is a conversion, not a photocopy. It recovers the content and structure — text, headings, lists, tables, reading order — and lays it out cleanly in Word. It does not reproduce the original page design pixel for pixel, because a Word document re-flows and a PDF does not. Anyone promising both is promising something the two formats cannot both be.
Tagged PDFs convert far better, and it says which yours is. The verdict above is read out of your actual file, not guessed from its size.
Scans need OCR, which this does not do. If a PDF is a photograph of a page there is no text in it to recover, and the page says so rather than handing you an empty document.
Word → PDF re-typesets the document. Real, selectable, searchable vector text — not a picture — but the lines break where this typesetter breaks them, not exactly where Word would. Matching Word's pagination needs Word's own fonts, which are not free to redistribute.
What it does not carry across: headers and footers, footnotes, comments, tracked changes, embedded images, columns, and text boxes. Those are dropped rather than mangled.
It learns locally, and only locally. Correcting the heading detection stores a number in this browser, against the software that produced the PDF. Nothing is collected, because nothing is sent anywhere — which is also why it cannot learn from anyone else's documents.
It downloads about 700 KB the first time, to read PDFs (pdf.js) and write them (pdf-lib). The Word half needs nothing extra: a .docx is a ZIP of XML, and the browser already has a ZIP engine.