Convert PDF to Word
Convert a PDF to an editable .docx Word document that keeps its exact look — processed entirely on your device, never uploaded.
- Free — No signup required
- Instant — Results update as you type
- Private — Processed in your browser
Editing a PDF usually means going back to the document that produced it, and when that document is gone, the PDF is all you have. This converter rebuilds a PDF as a .docx in which every line of text is real, editable Word text sitting at its original position, over an image of everything that was not text. It handles born-digital PDFs directly and falls back to optical character recognition (OCR) for scans, all in your browser.
How it works
Choose a PDF and press Convert to Word. For each page, the tool, using pdf.js:
- Extracts every text item with its position, font size and font name, and walks the page's drawing operators to find the fill colour in effect when each item was painted. Bold and italic are inferred from the font name (a
Helvetica-BoldObliqueglyph run becomes bold italic), and the PDF font is mapped to a comparable Word font. - Groups items into lines and renders the page to a canvas at 2× resolution.
- Erases each text line from that canvas by filling its box with the surrounding background colour, leaving logos, rules, table borders, photos and coloured panels intact. If nothing but flat colour remains, no image is stored at all.
- Builds the page with the
docxlibrary: a page of the same size with zero margins, the knocked-out canvas as a PNG anchored behind the text, and one absolutely positioned text frame per line carrying the original size, weight, style and colour. Because Word substitutes fonts, each line's characters are stretched or condensed slightly so it spans the same width as in the PDF.
The result downloads as your file's name with a .docx extension, with a count of the lines reconstructed.
Scanned and image-only PDFs
When a page has no embedded text, the tool switches to OCR with tesseract.js, the WebAssembly build of the Tesseract engine. On first use it downloads the engine and the English language data, around 15 MB, from a content delivery network and caches them; your document is not part of that download and never leaves your device. Recognised words are placed at the positions Tesseract reports, with bold and italic taken from its layout analysis, and the same knock-out and background image are applied.
Progress is shown per page because OCR is slow. Accuracy depends on the scan: 300 DPI, straight, high-contrast pages recognise well; phone photos, faxes, skewed or low-resolution scans produce errors that you will need to proofread. Only English is recognised; text in other languages will come out garbled.
Editing the result
The document is a faithful layout, not a flowing one. Each line lives in its own frame, so typing more into a line extends that frame rather than pushing later lines down, and deleting a paragraph leaves a gap. This suits correcting a figure, a date or a name on a statement, invoice or certificate, and is awkward for rewriting an essay. For prose you intend to restructure, extract the text (see below) and paste it into a fresh document instead.
Tables keep their appearance because the borders live in the background image and the cell text lies over them, but they are not Word tables. Hyperlinks, bookmarks and form fields are not carried across. Rotated or vertical text is placed at its anchor point but not rotated.
Command-line equivalent
For plain text with the original line structure, poppler's pdftotext is fast and dependable, and OCR of a scan can be done with Tesseract directly:
pdftotext -layout statement.pdf statement.txt
tesseract scan-page1.png output --dpi 300 pdf # searchable PDF from an image
Learn more
Batch extraction and OCR across a folder of files are shell jobs; shell basics covers the loops and text processing with grep, sed and awk shows how to pull structured fields out of the resulting text. Convert the edited document back with Word to PDF, or render pages as pictures with PDF to image if you only need a visual copy.
FAQ
- Will the original formatting be preserved?
- Yes — every line stays at its original position, size, weight and color and is fully editable, while the page's graphics (logos, tables, rules, images) stay as a background image. Convert the result back to PDF and it looks like the original.
- Does this work on scanned or image-only PDFs?
- Yes — when a PDF has no embedded text, the tool falls back to OCR (optical character recognition) running entirely in your browser, placing the recognized text back at its original position on the page. The recognition engine (~15 MB) downloads on first use and is cached afterward; accuracy depends on the scan quality.
- Is my PDF uploaded to a server?
- No. The conversion — including OCR — happens entirely in your browser; your PDF never leaves your device.