Site Logo

Greek · Image to Text

Greek from the cut stone to the printed page

Greek has been written for some twenty-seven centuries, and almost nothing that a modern reader relies on was there at the start: no spaces between words, no lower case, no accents, and for three hundred years no agreement about what the letters meant. Each domain below takes one stage of that history — a stone, a papyrus, a parchment codex, a minuscule book, a printed page — and sets out how to read it and how to check what comes back.

Choose the hand or document type

Transcribing Greek across the centuries

What survives is unusually varied. Tens of thousands of inscriptions record laws, treaties, accounts, dedications and epitaphs; the rubbish mounds and cemeteries of Egypt have given up papyrus by the hundred thousand, from tax registers and private letters to lost tragedies; the great parchment Bibles of the fourth and fifth centuries stand at the head of a manuscript tradition that runs unbroken to 1453; and after the fall of Constantinople the same texts were copied by emigre scribes and set in type at Venice, Paris and Basel.

The writing changed with every one of those materials. Archaic cities each cut their own alphabet and set the letters right to left, boustrophedon, or on a stoichedon grid; a reed pen on papyrus produced both the separated letters of a book hand and a rapid, ligatured documentary cursive; the codex brought biblical majuscule, with divine names contracted under a stroke; minuscule arrived about 800 and filled the page with ligatures, suspensions and, at last, accents and breathings; and the first Greek types copied those very ligatures, so that a printed book of 1500 still reads like a manuscript.

The people working on this material are epigraphers preparing corpora, papyrologists editing documents and literary fragments, textual critics collating biblical and classical codices, Byzantinists reading scholia and colophons, and rare-book librarians describing annotated early editions. Each page below explains what its hand looks like, why ordinary OCR fails on it, and what to tell a model before you trust the transcription.

Not sure which hand you have?

Upload a page, pick the closest domain, and test the prompt on one page before running the rest. You can switch domains at any time.