OCR (optical character recognition) reads a scan or a photograph of a printed page and gives you the text. Tibetan is hard for it: stacked letters, woodblock prints and handwritten scripts all need dedicated models.
The tools below are dedicated to Tibetan or built from Tibetan-trained models. Results vary a lot with the print and the scan quality, so plan to proofread.
A desktop app that runs Tibetan OCR on batches of images, including PDFs, and exports plain text or PageXML. Developed by Eric Werner for BDRC; installers for Windows and macOS. It works best with Uchen scripts.
A community-run guide to using Tibetan on digital devices, with pages on keyboards, fonts, dictionaries, OCR, converting old Tibetan text and more. Still under development.
BDRC's page on Hugging Face, "making Buddhist texts available to everyone using AI (OCR, search, etc.), focusing primarily on Tibetan". Hosts open OCR models and datasets.
Scripts that run Tibetan OCR on images and PDFs through Google Drive and write the result as text and Word files. Includes a quick-snip shortcut for macOS. Needs your own Google API credentials.
Missing a tool? Suggest it. See also all tools.
