The complete OCR toolkit for Linux — turn scanned PDFs and images into searchable, editable documents.
BigOcrPDF is a powerful, all-in-one OCR application that adds searchable text layers to scanned PDFs, extracts text from images, and provides a full-featured PDF editor — all from a modern, native Linux interface.
- AI-Powered OCR — Uses RapidOCR PP-OCRv5 with OpenVINO hardware acceleration for fast, accurate text recognition across 130+ languages
- Edit, Merge & Organize PDFs — Reorder pages, rotate, delete, and combine multiple PDFs and images into a single document
- Smart Preprocessing — Automatic perspective correction, deskew, dewarping, and illumination normalization — even photos of documents come out clean
- Multiple Export Formats — Searchable PDF, PDF/A-2b archival, plain text, and ODF/ODT with layout-aware formatting
- Screen Capture OCR — Select any region on screen and instantly extract text
- Batch Processing — Process dozens of files at once with checkpoint/resume support
- File Manager Integration — Right-click any PDF or image to OCR it directly
Manage your documents before and after OCR — no need for a separate tool.
- Drag-and-drop page reordering with thumbnail previews
- Rotate & flip pages — left, right, horizontal, and vertical
- Delete pages you don't need
- Merge files — combine pages from multiple PDFs and images into one document
- Create PDFs from images — import JPEG, PNG, TIFF, WebP, RAW photos, and more
- EXIF-aware import — automatically applies correct orientation from camera metadata
- Zoom control — 50% to 200% thumbnail scaling with keyboard shortcuts
- Select pages for OCR — choose exactly which pages to process
- Context menu — right-click any page to save as image or PDF
- Compress PDF — reduce file size with configurable quality and DPI
- Split PDF — by page count or target file size
- Undo support — revert page operations with Ctrl+Z
- Window size persistence — remembers your preferred dimensions
State-of-the-art text recognition powered by deep learning.
- RapidOCR PP-OCRv5 models with OpenVINO inference (ONNX fallback)
- 130+ languages across 12 script families: Latin, Chinese, Japanese, Korean, Arabic, Cyrillic, Greek, Devanagari, Tamil, Telugu, Thai, and more
- 4 precision levels — tune the trade-off between capturing hard-to-read text (tolerates more false positives) and strict recognition (avoids false positives but may miss low-legibility text)
- Parallel processing — multi-core batch OCR with automatic worker scaling
- Invisible text layer — preserves original page appearance while adding searchable text
- Smart detection — auto-identifies image-only vs. mixed-content PDFs
- Re-OCR support — replace existing text layers with improved recognition
- Right-to-left text — full BiDi support for Arabic and Hebrew via
fribidi
Automatically clean up scans and photos before OCR for maximum accuracy.
- Perspective correction — 6-mode cascade that straightens photographed documents
- Auto deskew — fixes tilted scans using morphological analysis + Hough transform
- Baseline dewarp — per-line polynomial fitting to flatten curved text
- Orientation detection — auto-correct 90°/180°/270° rotations
- Illumination normalization — even out uneven lighting
- Scanner effect — LAB-space background normalization
- Denoising — bilateral filter and Non-Local Means
- Enhance embedded images — apply corrections to images inside mixed-content pages
- All toggles individually controllable from educational settings dialogs with visual illustrations
Get your text out in the format you need.
| Format | Description |
|---|---|
| Searchable PDF | Original pages with invisible OCR text layer |
| PDF/A-2b | ISO archival standard with metadata injection (preserves original images) |
| Custom Quality PDF | Choose JPEG quality: 30%, 50%, 70%, 85%, or 95% |
| Black & White (JBIG2) | Pure black-and-white output using JBIG2 — the most compact format for text-only documents |
| Plain Text (.txt) | Extracted text from all pages |
| ODF/ODT |
4 modes: formatted + images, images + simple text, formatted text only, or plain text (experimental — formatting quality may vary) |
ODF export includes layout analysis: automatic paragraph/heading detection, table detection, image embedding, and proper page breaks. Note: ODF/ODT export is experimental and formatting results may not always be accurate.
Extract text from anything on your screen.
- Region capture — select an area and get the text instantly
- Works with: Spectacle (KDE), GNOME Screenshot, Flameshot
- Open any image — JPEG, PNG, WebP, TIFF, RAW formats (CR2, DNG, NEF, ARW, and more)
- Copy to clipboard with one click
- Standalone mode — run
bigocrimagefor a dedicated image OCR window
Handle large workloads efficiently.
- Multi-file queue — add files via drag-and-drop or file chooser, with grid and list views
- File information — right-click any file to view PDF metadata, fonts, images, and attachments
- Checkpoint/resume — interrupted sessions automatically resume on next launch
- Processing history — tracks file sizes, page counts, processing time, and success/failure
- Cancel anytime with clean cleanup
- Auto-split output — configurable maximum file size (10MB–100MB)
- Results page with per-file statistics, text viewer, and export actions
git clone https://github.com/biglinux/bigocrpdf.git
cd bigocrpdf
pip install -e .| Package | Purpose |
|---|---|
python >= 3.10 |
Runtime |
gtk4, libadwaita |
User interface |
python-rapidocr-pp-ocrv5 |
OCR engine |
python-rapidocr-openvino |
Hardware-accelerated inference |
poppler-utils |
PDF image extraction (pdfimages, pdfinfo) |
ghostscript |
PDF/A-2b conversion |
python-opencv |
Image preprocessing |
python-numpy |
Array operations |
python-pillow |
Image format support |
python-odfpy |
ODF/ODT export |
fribidi |
BiDi text reordering (Arabic, Hebrew) |
bigocrpdf # PDF OCR interface
bigocrimage # Image OCR windowbigocrpdf [OPTIONS] [FILES...]
Options:
-v, --version Show version and exit
-d, --debug Enable debug logging
--verbose Verbose output
--image-mode Launch in image OCR mode
FILES PDF or image files to open
- Right-click a PDF → Recognize text in scanned PDF (OCR)
- Right-click an image → Extract text from image (OCR)
- KDE Dolphin context menu integration included
Press Print Screen → select a region → export to Extract text from image (OCR).
- GTK4 + Libadwaita — clean, modern design following GNOME Human Interface Guidelines
- Multi-page wizard — Settings → Processing → Results
- Educational dialogs — image corrections, output, and advanced settings with SVG illustrations explaining each option
- Grid / List view toggle — switch between compact grid and detailed list in the file queue
- Context menus — right-click files in the queue or pages in the editor for quick actions
- Toast notifications — non-intrusive status feedback
- Before/After comparison — track file size changes after OCR
- Window size persistence — remembers your preferred dimensions for all windows
- Keyboard shortcuts — comprehensive shortcuts for all major actions
- 28 UI languages — Bulgarian, Chinese, Czech, Croatian, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hebrew, Hungarian, Icelandic, Italian, Japanese, Korean, Norwegian, Polish, Portuguese, Romanian, Russian, Slovak, Spanish, Swedish, Turkish, Ukrainian
graph TD
A[bigocrpdf] --> B[Application Layer]
A --> C[Services Layer]
A --> D[UI Layer]
A --> E[Utils Layer]
B --> B1[application.py<br/>Adw.Application entry point]
B --> B2[window.py<br/>Main PDF OCR window]
B --> B3[config.py<br/>Constants & configuration]
C --> C1[processor.py<br/>OCR engine interface]
C --> C2[screen_capture.py<br/>Screen capture & image OCR]
C --> C3[export_service.py<br/>PDF / Text / ODF export]
C --> C4[contour_analysis.py<br/>Document contour detection]
C --> C5[perspective_correction.py<br/>Geometric correction]
C --> C6[rapidocr_service/]
C6 --> C6a[engine.py — Singleton OCR engine]
C6 --> C6b[ocr_worker.py — Subprocess worker]
C6 --> C6c[preprocessor.py — Image pipeline]
C6 --> C6d[rotation.py — Orientation detection]
D --> D1[image_ocr_window.py<br/>Standalone image OCR]
D --> D2[settings_page.py<br/>OCR settings]
D --> D3[conclusion_page.py<br/>Results & export]
D --> D4[pdf_editor/<br/>PDF page editor]
E --> E1[odf_exporter.py<br/>ODF document generation]
E --> E2[layout_analyzer.py<br/>Document structure detection]
E --> E3[checkpoint_manager.py<br/>Session resume support]
style A fill:#4A86CF,color:#fff
style C6 fill:#3776AB,color:#fff
- 311 automated tests covering OCR pipeline, PDF operations, export, preprocessing, editor logic, and utilities
- Tested with Python 3.10 through 3.14 — supports the latest Python release
- 100% i18n coverage — all 28 languages fully translated (604 strings each)
- Ruff-enforced code style and linting
- WCAG 2.1 Level AA accessibility considerations