Collating ten manuscripts of a work by hand takes months. Given their readings, this tool finds every point where the copies differ, lists each variant under its manuscript, and drafts the apparatus. The editor decides every reading; the tool never does.
Wherever a witness differs from the base text, or the OCR is unsure, the reading is shown beside the photograph of that spot on the leaf.
A low score sends the place to the editor. The OCR is never allowed to "correct" a scribe's reading into the familiar printed one.
Each accepted or rejected reading is stored with the editor's note, so the edition can be reviewed and extended later.
The first test is the Viṣṇusahasranāma from six Devanāgarī manuscripts in our collection: a short, well-known text with real variation between copies. Five copies of the Devīmāhātmya and three each of the Gītā, Raghuvaṃśa and Gītagovinda come next.
| Step | When | Status |
|---|---|---|
| OCR able to read the six hands | December 2026 | Coming soon |
| Collation and variant list | January 2027 | Works today (engine in the workbench) |
| Draft apparatus and TEI-XML | January 2027 | Coming soon |
| Stemma suggestion | February 2027 | Coming soon |
The same collation engine also serves the collated editions of the IKS Texts, where printed editions and e-texts are compared.