Scanned PDF to Word

Why a normal PDF to Word converter gives you a picture

If your PDF came out of Word or Google Docs in the first place, it already has real text inside, and our regular PDF to Word converter will keep the original layout much better than OCR can.

Most PDF to Word converters, ours included, work by reading the text that's already stored inside the PDF. That's great for a file exported from Word or Google Docs. A scanner doesn't store any text, though. It stores an image of each page. So the converter finds nothing to read, shrugs, and drops the image into the Word file.

You open the .docx, click on a paragraph, and the whole page lights up as one picture. I've had that exact moment with a lease agreement a landlord sent over, and it's maddening. The text is right there. You can see it. You just can't touch it.

Fixing that needs OCR (optical character recognition) to happen first. Something has to look at the pixels and decide that this shape is an "a" and that one is a "7". Only then is there text to put into Word.

What this tool does with your scan

Every page goes through Tesseract, the open source OCR engine that most serious OCR software is built on. Before it reads anything, the page is straightened, because a scan that's tilted by a degree or two loses a surprising number of words. Pages that already contain real text are left alone and their text is used as it is.

Then comes the part I spent the most time on. Raw OCR output is a pile of lines. If you paste it straight into Word you get a ragged mess, with every line ending in a hard break and every hyphenated word split in two. The converter looks at the gaps between lines and the size of the text to rebuild proper paragraphs, joins words that were broken across a line ending, and turns large lines like a letterhead or a section title into real Word headings. Each page of the scan starts on a new page in the document.

What you download is a plain, clean .docx. Normal styles, no text boxes floating on top of an image, nothing hidden. You can change the font across the whole thing in one go, run spell check, track changes, or paste sections into another document without dragging junk formatting along.

Picking the right language matters more than you'd think

Tesseract reads much better when it knows what language to expect. Set it to English and feed it a Spanish contract, and you'll still get most of the words, but the accents go missing and "año" may turn into something embarrassing. German loses its umlauts. French loses its cédilles. For Japanese, Chinese, Korean, Arabic, Hindi or Thai, the wrong setting gives you near garbage, because those scripts need their own trained models.

There are 20 languages in the list. If you're on the Spanish, German, French, Portuguese, Japanese or Chinese version of this page, the matching language is already selected for you. Documents that mix two languages work best when you pick the one that covers most of the text.

Getting a clean result

OCR quality depends almost entirely on the scan. A few habits make a big difference:

• Scan at 300 DPI. Lower than 200 and small print starts to blur into blobs the engine can't separate.
• Black and white or greyscale is fine. Colour doesn't help the OCR and makes the file bigger.
• Flatten the page. Book spines curve the lines near the middle, and curved lines read badly.
• Phone photos work if the page fills the frame and you shoot straight down, not at an angle.
• Faxes and photocopies of photocopies will work, slowly and with more mistakes. Expect to proofread.

Even on a good scan, read the result before you send it anywhere. OCR confuses 0 and O, 1 and l, rn and m. Numbers in invoices and contracts deserve a second look.

Where it works well, and where it doesn't

If you only need to search or copy from a scan and want the pages to look exactly the same as before, the OCR PDF tool adds a hidden text layer to the PDF instead of rebuilding it as a Word file.

It's at its best on text-heavy pages: letters, contracts, reports, meeting minutes, old typed manuscripts, court documents, book chapters. Anything where the job is getting the words back so you can edit them.

I'll be honest about the limits. Tables come out as lines of text rather than a Word table, so a dense price list will need tidying. Two-column layouts are usually read in the right order, column by column, but a page with sidebars and callout boxes can come out shuffled. Photos and logos on the page are not carried over. And handwriting is mostly a lost cause; Tesseract is trained on print, and cursive notes in the margin will either vanish or turn into nonsense.

Paid tools like Adobe Acrobat try to rebuild the exact look of the page, fonts and all. Sometimes that's impressive. Often it produces a Word file full of positioned text boxes that falls apart the moment you add a sentence. I went the other way on purpose: get the words right, keep the structure simple, and let you format it the way you want.

What happens to your file

Password-protected scans can't be read. If yours has one and you're allowed to open it, remove the password first and then convert the unlocked copy.

OCR is heavy work, so it runs on our server rather than in your browser. Your PDF is uploaded over HTTPS, processed in a queue, and deleted as soon as the conversion finishes. The Word file stays available long enough for you to download it and is then removed automatically. Nobody reads your documents, and nothing is used to train anything.

There's no account, no email gate and no watermark stamped across your text. The limit is 100 pages per file, which keeps a single conversion under a few minutes. Most of the time it's well under that; a ten page letter usually takes less than a minute.

A few jobs people use it for

A small business owner had a supplier contract that only existed as a signed scan, and needed to send back a marked up version. Converting it took a minute; fixing the three clauses took ten.

A student had 40 pages of a library book photocopied for a dissertation and wanted to pull quotes without retyping them. A family historian turned a typed 1960s family memoir into a document the cousins could add to. An office manager rescued a policy handbook whose original Word file was lost years ago, with only a printed copy left in a binder.

None of those people needed a perfect replica of the page. They needed the words back. That's what this does.