Sanskrit · Image to Text
One language, a different alphabet in every region
Sanskrit has no script of its own. Wherever it was copied, it was written in the local descendant of Brāhmī, so the same verse looks entirely different in Kashmir, in Nepal, in the Tamil country and in the Deccan — and the modern habit of printing everything in Devanāgarī hides that. Each domain below takes one of those regional scripts and the material it was written on, and explains what has to be told to a model before a transcription can be trusted.
Choose the hand or document type
Siddham Buddhist Manuscriptsc. 600–1200Transcribe Siddhaṃ Sanskrit from palm leaves, Dunhuang scrolls and Japanese sutra prints: conjuncts, vowel signs, anusvāra and visarga, checked folio by folio.
Śāradā Manuscripts of Kashmir and the Northwestc. 900–1700Transcribe Kashmiri Śāradā manuscripts on birch bark and paper: angular akṣaras, dense conjuncts, Vedic accents and marginal glosses, checked against the folio.
Newa and Nepalese Sanskrit Manuscriptsc. 1000–1900Transcribe Nepalese Sanskrit manuscripts in Bhujimol, Nepālākṣara and Newa: palm leaves with string holes, dated colophons and Newar-language glosses.
Grantha Sanskrit Manuscripts of the Tamil Regionc. 1100–1900Transcribe Grantha palm-leaf manuscripts from the Tamil country: incised rounded letters, stacked conjuncts, no word spaces and string holes through the text.
Nandinagari Sanskrit Manuscriptsc. 1000–1900Transcribe Nandinagari: the Deccan cousin of Devanagari, on palm leaf, paper and Vijayanagara copper plates, with its own letterforms and conjuncts.
Transcribing Sanskrit across the centuries
What survives is spread over two thousand kilometres and fifteen centuries. Buddhist dhāraṇīs travelled east in Siddham and were kept in China and Japan long after they had gone from India; Kashmir wrote its philosophy and its chronicles in Śāradā on birch bark and paper; the Kathmandu valley kept an unbroken run of dated manuscripts in Newa hands; the Tamil country wrote Sanskrit in Grantha on incised palm leaf; and the maṭhas of the Deccan copied the Vedas and the Dvaita commentaries in Nandinagari.
The material dictates the letters. A stylus cannot cut a straight horizontal into palm leaf without splitting the fibre, so the southern scripts curve and loop, while a reed pen on birch bark or paper gives the northern hands their angular tops and long verticals. All of them write continuously, with sandhi joining word to word and no spaces at all, stack consonant clusters into ligatures that must be learned one by one, and close their units with a daṇḍa.
Those working on it are editors preparing critical editions from untranscribed witnesses, Vedic and Āgamic specialists reading ritual manuals, cataloguers describing bundles held at title level across South Asia, and digitisation projects deciding which script to encode a text in. Each page sets out what its hand looks like, why generic OCR fails on it, and what to check first in 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.