Repository navigation
HeOCR 0.1.0 for macOS and Windows. Drop in a PDF, a Word file (DOCX) or an image, and get its Hebrew text out, with its headings, paragraphs and tables. Copy it, or save it as plain text, Markdown or JSON.
Fully offline. Your files never leave your computer. The app makes no network connections at all: no telemetry, no analytics, no update checks, no downloads.
Download
macOS
HeOCR-0.1.0-macos.dmg: macOS 13 or later, one app for Apple silicon and Intel.- Signed with an Apple Developer ID (Shay Palachy Affek, 8X657D2YAU) and notarized by Apple.
- sha256
7e963f8b818e54541bcf72d6a21d72143bc1bca63b97611a857972d516fdc919
Windows
HeOCR-0.1.0-windows-x64.msi: Windows 10 or 11, 64-bit.*- sha256
4a4ada41b6a035f0a3f88332f5cec55620740e25ae8f8b19dd21915bfebb44ae - Built by GitHub Actions from this release's commit (
249ca04, CI run). Before it was attached, it was installed and smoke-tested on a clean Windows runner.
* Windows: unsigned. Windows SmartScreen will warn; choose More info → Run anyway. A signed build follows soon.
Inside
- heocr 0.1.2, pinned as
heocr==0.1.2from PyPI (tagv0.1.2,ac1ad20): the version that passed heocr's release gate, plus its fixes. - Tesseract 5.5.3, built from source without network code.
- OCR models:
hebfrom tessdata_best 4.1.0 andengfrom tessdata_fast 4.1.0, the same files heocr's release gate ran with. - Licences: everything is under a permissive licence; see THIRD_PARTY_LICENSES.md.
- Source: this repository at
249ca04.
Good to know
- OCR text is a reading, not a verified transcript: check names and numbers.
- Older Word files (.doc), RTF and OpenDocument files need LibreOffice to convert, which the app does not include. Save them as DOCX or PDF first.