PDF to Word converter

How a PDF becomes a Word document

A PDF does not know about paragraphs. It stores each piece of text at a position on the page, in a font and a size, and that is all. To make an editable document, the converter reads the page the way a person does: it groups words into lines by their baseline, joins lines into paragraphs, and then looks for structure.

• Headings: lines larger or bolder than the body text become Heading 1, 2 or 3, so the navigation pane in Word works.
• Tables: ruled grids and neatly aligned columns of text become real Word tables, with merged cells and shading.
• Lists: lines that start with bullets or numbers become Word lists that renumber themselves.
• Columns: text set in two or three columns stays in columns instead of being read across.
• Headers and footers: lines repeated at the top or bottom of every page, such as page numbers, move into the Word header and footer.

Fonts that every computer has, such as Arial, Calibri or Times New Roman, keep their names. Others are replaced by Arial, Times New Roman or Courier New, matched by style, so the document looks the same on any computer that opens it.

What we measured

Documents that were made on a computer convert in a second or two, because their text is already there. Scans have to be read letter by letter first, which took about 12 seconds a page. Word files are small because text takes little space; with pictures, they grow with the photos they carry.

The switches

Every part of the rebuilding can be turned off when it gets in the way. Switch off Find tables if text in a layout was turned into a table you did not want, or Keep columns to get one simple column to edit. Keep line breaks, off by default, ends every line where it ended in the PDF: the result looks closer to the original, but editing a paragraph then means fixing every line by hand. Pages to convert takes ranges such as 1-3, 5 for long documents.

Scanned pages

By default, text recognition runs only on pages that have no text of their own, so a document with a few scanned pages in it is handled page by page. Choose On every page for a PDF whose text layer is wrong or garbled, and pick up to three languages that appear in the scan. Recognition reads printed text, not handwriting. To make a scan searchable while keeping it exactly as it looks, image to text also makes a searchable PDF.

What does not come across

• Form fields are not reproduced. Fill in a PDF form in a PDF reader instead.
• Charts and drawings made of lines and shapes are not turned into Word drawings. Photos and other pictures are kept.
• Very designed pages, such as magazine layouts, posters and text placed over pictures, come out as simpler documents. The text is all there, but the arrangement is not.

If you only need the words, PDF to text is faster and gives plain text with no layout at all.

If something goes wrong

• It says the PDF is password protected. Unlock it first with unlock PDF, if you have the right to.
• The text comes out as nonsense characters. Some PDFs use fonts with no proper mapping to letters. Convert again with text recognition set to On every page.
• Arabic or Hebrew reads backwards. Set the text direction to Always right to left.

How we tested. Three test PDFs: a 10 page report exported from Word, an 8 page brochure with four photos, and the same report drawn as a 300 DPI scan without any text. Converted with this tool's code and default settings in Chrome.