Easy Tibetan Copy

Copy, paste and search Tibetan PDFs — at last.

Drop a PDF, Word or RTF file and get correct Unicode back. Everything runs in your browser. Your file never leaves your device.

   

---

Why

Many Tibetan PDFs, Word and RTF files look perfect on screen but fall apart the moment you select the text: copy-paste returns gibberish, search finds nothing, and the characters come out in some pre-Unicode legacy encoding. Easy Tibetan Copy repairs them so the Tibetan comes out as correct, searchable Unicode — right in the browser, with no server and nothing uploaded.

What it does

🩹 Fix the PDF	Repairs the embedded font's /ToUnicode map so copy-paste, search and extraction return correct Unicode, then hands back a fixed PDF to download. Pre-Unicode legacy Tibetan fonts (TibetanChogyal, Ededris/Dedris, …) are handled automatically — no toggle, no extra step.
📄 Extract text	Pulls clean Unicode out of the PDF, optionally taking only odd or only even pages (handy for pecha-style books printed two-up). The preview and the exported Word .docx keep the original font sizes, bold/italic and paragraph flow; a plain .txt is also available.
🔁 Convert Word / RTF	Drop a .docx or .rtf whose Tibetan is in a legacy font and get the same file back with only the Tibetan runs swapped to Unicode — all other formatting preserved. Old binary .doc is detected and you're asked to re-save as .docx first.

Privacy

It is 100% client-side. Your document is processed entirely in your browser and is never uploaded — nothing leaves your device. The only things fetched over the network are the runtime assets (the Pyodide engine, the pdf-cmap-fix wheel and jszip), all cached after the first visit. The whole app is static and hostable on GitHub Pages.

How it works

There are two independent pipelines, picked automatically by file type:

web/index.html        UI (+ an import map for jszip)
web/app.js            state machine — routes by file type (pdf / docx / rtf / doc)
web/worker.js         Web Worker: Pyodide + pdf-cmap-fix + python-docx  (PDF path)
web/sw.js             service worker — caches Pyodide + the wheel for fast repeat loads
web/wheels/           the pdf-cmap-fix wheel (built by scripts/build-wheel.sh; gitignored)
web/vendor/tibetan-ansi-to-unicode/   vendored JS converter             (Word/RTF path)

PDF path — the UI hands the bytes to a Web Worker running Pyodide 0.29.4 (which bundles PyMuPDF 1.26.3 + fonttools). The worker installs the pure-Python pdf-cmap-fix wheel and python-docx, repairs/extracts in memory, and returns the result for download.

Word / RTF path — no Pyodide. The app lazy-imports the vendored tibetan-ansi-to-unicode module, converts the document in place, and offers the converted file for download.

Run it locally

./scripts/build-wheel.sh          # builds web/wheels/pdf_cmap_fix-…whl (needs python3 + pip + git)
cd web && python3 -m http.server  # then open http://localhost:8000

Any static file server works — there is no build step and no backend.

Deployment (GitHub Pages)

Pushing to main triggers .github/workflows/deploy-pages.yml, which builds the wheel and deploys web/ to GitHub Pages. One-time setup: in the repo settings, enable Pages with the “GitHub Actions” source.

Notes & limits

• Large files run in the tab's memory. Files over ~20 MB trigger a warning (the tab can run out of memory) but can still be processed — a desktop is recommended for big PDFs. Typical few-MB Tibetan PDFs are comfortable, even on mobile.
• No Markdown export in the browser build: pymupdf4llm needs a newer PyMuPDF than Pyodide bundles. Extraction instead repairs the PDF, reads PyMuPDF's structured text, and preserves formatting into the .docx (plus a plain .txt).
• Legacy-font coverage depends on the font being known to pdf-cmap-fix. Report fonts that don't convert upstream at OpenPecha/pdf-cmap-fix.

Credits & license

Wraps OpenPecha/pdf-cmap-fix (MIT) for PDFs and vendors tibetan-ansi-to-unicode for the Word/RTF conversion. The preview and exported .docx use the Jomolhari Tibetan font (SIL Open Font License), bundled at web/fonts/. See DECISIONS.md for the reasoning behind the design. This project is released under the MIT License.

_{Built for the Tibetan text community · buda-base.github.io/easy-tibetan-copy}