Converting Greek (and other non-Latin) PDFs to Word

Last reviewed: 2026-10-25

Greek is the second script our visitors use most, so we tested it properly. The short version: a normal Greek PDF converts perfectly; a scanned one is now good but not perfect; and a few Greek PDFs carry a font trap worth knowing about.

Greek PDFs with real text: no problem

If you can select individual Greek words in the PDF, the text is stored as Unicode characters and conversion simply copies them. Our Greek test form, with accents, capitals and diaeresis (προϋπόθεση, αϋπνία, Μαΐου), came out 116 words out of 116, in order, with its table intact.

Scanned Greek: why it used to fail

OCR software has to be told which languages to expect, and only uses the language packs that are installed. OCRmyPDF assumes English unless told otherwise. Until 2026-10-25 our server read every scan as English, and our scanned Greek form came back as Latin-letter gibberish: about 3% of words right.

The server now reads scans with English and Greek together. On the same scanned form, 93.6% of words now come out right. The remaining mistakes were in two places:

• The large bold title was misread (ΑΙΤΗΣΗ ΓΙΑ ΒΕΒΑΙΩΣΗ ΣΠΟΥΔΩΝ came out as nonsense mixing Greek and Latin letters). Very large or decorative type is harder than body text.
• Letters with diaeresis were sometimes read with a different mark, for example ϋ as ὐ in προϋπόθεση. A Find & Replace in Word fixes these in one go once you spot the pattern.

One table label and its value were also joined without a space. Body paragraphs were otherwise correct.

Convert a PDF now → Free, no sign-up, files deleted as soon as you download the result

The ∆ trap in Greek PDFs

Some Greek PDFs, especially older public-sector forms, use a font that draws the capital Delta with the mathematical “increment” sign ∆ (Unicode U+2206) instead of the Greek letter Δ (U+0394). They look identical. After conversion the text looks right, but searching for “ΔΗΜΟΚΡΑΤΙΑ” will not find “∆ΗΜΟΚΡΑΤΙΑ”, and spell-checkers mark the word as wrong.

Our converter treats ∆ as a letter when it checks for glued-together words, so it does not break those words apart. It does not change the character itself, because we cannot be sure every ∆ in a document is a Delta. To fix it in Word: Ctrl+H, paste ∆ into “Find what”, type Δ in “Replace with”, Replace All.

Other signs of a broken font encoding

If a PDF looks fine but the converted text is random symbols or Latin letters in place of Greek, the PDF itself stores its characters with a private encoding and no map back to real letters. Copying text out of the PDF in any viewer gives the same garbage. No converter can read the text correctly from such a file. The workaround is to treat it as a scan: print it to an image-based PDF (or take screenshots) and convert that, so OCR reads the shapes instead.

Other scripts

• Text PDFs in any script (Cyrillic, Arabic, Hebrew, Chinese and so on) are converted as text, so the characters come through. Right-to-left layout and vertical text are another matter: we have not tested them and you should check the result carefully.
• Scanned PDFs are read with English and Greek only. Tesseract supports more than 100 languages, but each one adds processing time and may lower accuracy for the others, so we only installed the two our visitors need most. A scan in Russian, Arabic or Chinese will not be recognised correctly here. Latin-alphabet languages such as Spanish, French, German or Italian are mostly read correctly with English, though accented letters can be wrong.

Frequently asked questions

Can I convert a Greek PDF to Word?

Yes. Greek PDFs with real text converted word for word in our test. Scanned Greek PDFs are read with Greek OCR; our scanned test form reached 93.6% of words correct.

Why does search not find words with Δ in my converted file?

Some Greek PDFs use the mathematical sign ∆ (U+2206) in place of the Greek letter Δ (U+0394). Replace ∆ with Δ using Find and Replace in Word.

Can you OCR Russian, Arabic or Chinese scans?

Not at the moment. Scans are read with English and Greek language data only.