Persian · Image to Text
Persian from the first codices to the last farmans
New Persian has been written in the Arabic alphabet since the ninth century, and much of the difficulty comes from that borrowing: four extra consonants that lose their dots first, a script with no capitals to lean on, and hands that gave up the baseline altogether. Each domain below takes one of those hands and sets out what is on the page and how to check a transcription of it.
Choose the hand or document type
Early New Persian Kufic and Naskh Manuscriptsc. 900–1200Transcribe early New Persian in eastern Kufic and naskh: interlinear Qurʾan glosses, unwritten short vowels, and پ چ ژ گ hiding inside Arabic letters.
Persian Chancery Taʿliqc. 1200–1500Read Ilkhanid, Jalayirid and Aq Qoyunlu farmans in taʿliq: hanging lines, letters joined that should not join, seals and endorsements at every angle.
Classical Persian Nastaʿliq Literary Manuscriptsc. 1400–1700Transcribe Persian poetry in nastaʿliq: falling word groups, letters piled inside a word, ruled columns, diagonal hemistichs and margins full of glosses.
Persian Shikasta Nastaʿliq Correspondencec. 1650–1900Read shikasta letters, petitions and farmans: letters joined that should not join, dots dropped or displaced, rising lines and margins full of endorsements.
Indo-Persian Nastaʿliq Administrative and Literary Handsc. 1500–1900Transcribe Mughal and successor-state Persian: farmans with tughras and seals, revenue figures in siyāq, deeds, letters and court poetry in Indian nastaʿliq.
Transcribing Persian across the centuries
The surviving material is unusually varied. Early copies of the Qurʾan carry interlinear Persian versions under the Arabic; chancery archives preserve Ilkhanid and Timurid decrees, endowment deeds and letters; libraries hold thousands of illuminated dīvāns and epics; family papers run from Safavid petitions to Qajar correspondence; and in South Asia revenue records, chronicles and deeds were kept in Persian until the 1830s.
The hands changed with the job. A tenth-century copy uses an angular eastern Kufic or an early naskh with sparse pointing; the chancery answered with taʿliq, a hanging hand that piles words on a slope; nastaʿliq was codified around 1400 and made the page a set of falling word groups ruled into columns; shikasta then broke it for speed, joining letters that join nowhere else and dropping the dots entirely. In India the same nastaʿliq grew broader and flatter, and shared its documents with siyāq figures, seal impressions and endorsements in Indic scripts.
Those who work with it are editors collating poetry across dozens of manuscripts, cataloguers describing uncatalogued collections, historians reading decrees and revenue papers, and archivists opening boxes of Qajar and Mughal correspondence. Each page explains what its hand looks like, why ordinary OCR fails on it, and what to tell a model before you trust the result.
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.