PDF to Word

Why everyone else uploads your document to do this

Because it is the easy way. Converting on a server means running a full office suite on it, and that costs money per file, which is why every free PDF to Word converter caps you at one document and asks for an email at the second.

iLovePDF limits it to a single 15MB file whether you have an account or not, and licenses the conversion engine from a third party. Smallpdf will not publish a number for any of its limits. Both keep your file for a while afterwards.

This runs entirely on your machine. There is no server, so there is nothing to meter and no copy of your document anywhere.

How it works, and what that means for the result

Worth understanding, because it explains what you will get and why nobody can promise more.

A PDF does not contain paragraphs. It contains pieces of text, each with a position on the page, in whatever order the program that made it happened to write them. There are often no spaces either: the word "Balustrade" can arrive as "Bal", "ust" and "rade" as three separate items.

So the first job is working out which pieces belong on the same line, in what order, and where a gap is wide enough to be a space rather than just letter spacing. Get that wrong and you get the wall of jumbled words that most converters produce.

A .docx is then a zip file containing XML. That part is genuinely straightforward once the text is in order, so "PDF to Word needs a server" was never really true. The hard part was always the PDF end.

Tables are found by looking for columns

This is where nearly every free converter gives up, and it is why the output is usually unusable. Read the text in order and a six column timesheet arrives as "Monday 07:30 16:00 30m 8h 8.00" on one line, as a paragraph, with nothing to edit.

A PDF holds no table. It holds text at positions, and the lines around it, if there are any, are separate drawing instructions that know nothing about the text. The signal that isthere is alignment: cells in a column start at the same place on row after row, and ordinary prose never does that.

So the text on each line is split into cells at the wide gaps, and any run of consecutive lines sharing those positions becomes a real Word table, with the header row shaded and set to repeat across a page break. A run also breaks at a large white gap, so a summary block sitting under a grid stays a separate table instead of being swallowed into it.

If it gets a table wrong on your document, turn the switch off and every row comes through as a line of text instead.

The logo and the pictures come with it

Pictures are found by where they are placed on the page, then cut out of a 200dpi render of that page and embedded in the Word file at the size they were printed at. Cutting them from the render rather than pulling the raw objects out is deliberate: a PDF can store a bitmap in a dozen ways, and getting the colour space or the mask wrong gives you an inverted or a black logo. What is on the paper is what you get.

Anything covering most of the page is left out, because that is a scanned background rather than a logo, and lifting it just hands you the page back as a picture.

Headings are inferred from type size

A PDF has no styles, no heading levels and no structure. The only signal left about what is a heading is that somebody set it bigger than everything around it.

So lines noticeably larger than the body text become real Word headings, which means they show up in the navigation pane and can build an automatic contents page. The body size is taken as the median line size rather than the average, because one large title on the front page drags an average up far enough that nothing else qualifies.

A long line is never treated as a heading however large it is. That is a pull quote or an opening paragraph in bigger type, and marking it as a heading puts nonsense in the contents.

What does not come across

Said plainly, because a converter that oversells this is worse than one that explains it.

Exact fonts, colours and spacing are not reproduced: the document is rebuilt in Calibri with sensible sizes rather than made to look identical.Multiple columns of prose are read down the page rather than column by column.Headers and footers arrive as ordinary text at the top and bottom of the page rather than as real Word headers.

The structure above was worked out rather than read, so check the preview against your document before you rely on it. What you get is an editable document with the words, the tables, the pictures and the headings, instead of one you have to retype. For most of what people actually need, which is amending a contract, reusing a specification or lifting a method statement, that is the job.

If the result is empty, it is a scan

A scanned PDF is a photograph of paper. There is no text in it, only a picture of text, so there is nothing to convert and a Word file made from it would be blank.

This checks before doing anything and says so rather than handing you an empty document and letting you work it out. It is the same reason copying and pasting from that PDF gives you nothing.

Getting words out of a scan needs text recognition, which reads the shapes of the letters in the picture. That is a different job and a separate tool.

Contracts, specifications, and things you did not write

The PDF is read and the Word file is built on this page using your own machine. Nothing is uploaded, nothing is stored and there is no account, which matters given that the documents people convert are usually contracts, specifications and things they did not write.