dana

Hebrew PDF Copy

Fixes scrambled Hebrew and English text copied from PDF files, such as clinical summaries viewed in Acrobat.

Open pdfcopy.html in a browser (Edge or Chrome), then either:

• Paste a screenshot of the paragraph (Win+Shift+S, then Ctrl+V). The text is read with built-in optical character recognition.
• Load or drag the PDF file, then mark the area you want. The text is read directly from the file, so this is the most accurate option. In a scanned PDF (no text inside the file), the marked area is read from the page image, like a screenshot.

The corrected text is shown right to left and can be edited and copied. Words the tool is unsure of are shown in red.

Privacy

Everything runs inside the browser on your computer. No text, screenshots or PDFs are uploaded or saved, and no language models (LLMs) are used. The only network request is an anonymous usage counter (GoatCounter: a visit or a use is counted, never any content); the page’s security policy blocks every other request. The file works offline once downloaded.

Built with

• Tesseract.js with the Hebrew and English best_int models, embedded in the file (Apache License 2.0)
• pdf.js by Mozilla, embedded in the file (Apache License 2.0)
• The Hebrew PaddleOCR word recognizer by Rivok (Rivok/paddleocr-hebrew, mobile-word model, fine-tuned for Hebrew and English written together), a second reader, embedded in the file (Apache License 2.0)
• ONNX Runtime Web by Microsoft, which runs that model in the browser, embedded in the file (MIT License)

License

This project is licensed under Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0): free to use, share and adapt for non-commercial purposes, with attribution. Commercial use is not permitted. See LICENSE.

The embedded third-party components keep their own licenses (Apache License 2.0, MIT); see THIRD-PARTY-NOTICES.md.

Created by Dr. Tom Rabinowitz @ Dana-Dwek Children’s Hospital, Tel Aviv Sourasky University Medical Center

This site is open source. Improve this page.