Why a PDF is hard to turn back into a document
A Word file stores text as a flow: paragraphs, headings and lists that rewrap when you change the margins or the font. A PDF stores something else entirely. Each page is a set of instructions such as "draw these characters at this position in this font", with no notion of where a paragraph starts or which line belongs to which column. That is why a PDF looks identical everywhere and is so awkward to edit.
Converting PDF to Word therefore means reconstructing structure that the PDF never recorded. Any converter has to guess, and the two options on this page guess in different ways.
The browser rebuild, step by step
The default path runs entirely on your device, using the same PDF engine Firefox uses to display PDFs. For every page it collects the text pieces with their positions and font sizes, then works upward:
- Lines. Pieces sitting on the same baseline, within about 2 points of each other, are joined into one line.
- Body size. The font size that carries the most characters across the whole document is taken as normal body text.
- Headings. A line more than 1.25 times the body size becomes a heading; above 1.5 times it becomes Heading 1, otherwise Heading 2. A short bold line under 80 characters at body size or larger also becomes Heading 2.
- Paragraphs. A line that does not end with sentence punctuation such as a full stop, question mark, colon or closing quote, and has the same font size as the next line, is treated as continuing into it.
Because headings use Word's real heading styles, they appear in the navigation pane and can feed an automatic table of contents straight away. Font sizes are carried over, and bold survives on lines that are not headings. The page shows the number of pages and words it found, which is a quick sanity check that nothing went missing.
What this path deliberately does not attempt: images, tables, columns, colours, hyperlinks and the original font families. Pages follow each other as one continuous document, separated by an empty paragraph. For a contract or report you want to rewrite, that is usually what you need: clean text without floating boxes.
The server conversion and when it is worth it
The second option sends the file to our server, where LibreOffice opens the PDF in its word processor and saves it as a .docx. This keeps much more of the look: images, lines, text placement and page breaks stay close to the original.
The trade-off is how the text is stored. LibreOffice usually places text in separately positioned boxes, often one per line, so the page looks right but rewriting a paragraph means editing box by box. Choose the server path when you need a near copy to tweak a few words, such as correcting a date on a form or a figure in a flyer. Choose the browser path when you need to rework the text itself.
Without an account you get one server run a day for files up to 500 MB. Premium removes the daily limit and accepts files up to 4 GB. The uploaded PDF is deleted when the conversion finishes and the .docx expires about an hour after the job was created.
Fixing typical results
- List items merged into one paragraph. Bullet points without closing punctuation look like a continuing sentence to the paragraph rule. Put the cursor at each item and press Enter, or apply a list style afterwards.
- Words split by a hyphen and a space. If the PDF hyphenated words at line ends, the join produces "conver- sion". A find-and-replace for "- " (hyphen, space) catches most of them; review each match, since some are genuine dashes.
- Two-column text interleaved. The browser path reads text in the order the PDF stores it, which in multi-column layouts can jump between columns. The server conversion handles these pages better.
- Running headers and page numbers in the middle of the text. They are real text on every page, so they come through as short paragraphs. Delete them, or search for the repeated title.
- The conversion fails immediately. Password-protected files cannot be read. Remove the password with Unlock PDF first, as long as you know it.
Scanned documents
If the result is empty or nearly empty, the PDF is a scan: each page is a photo with no text in it. Run the file through OCR PDF to add a text layer, then convert the searchable copy here. Expect a rougher result than with a born-digital PDF, because recognised text has less consistent font sizes, which makes heading and paragraph detection less reliable.