How to OCR PDF Files on Linux with Tesseract

Tesseract installation is straightforward on all major Linux distributions. Installing language packs is equally important — without the right language data, Tesseract's accuracy drops significantly.

1. 1Install Tesseract on Ubuntu/Debian: `sudo apt install tesseract-ocr`
2. 2Install Tesseract on Fedora/RHEL: `sudo dnf install tesseract`
3. 3Install English language data (usually included): `sudo apt install tesseract-ocr-eng`
4. 4Install additional languages (e.g., French, German, Spanish): `sudo apt install tesseract-ocr-fra tesseract-ocr-deu tesseract-ocr-spa`
5. 5Verify installation and list available languages: `tesseract --version && tesseract --list-langs`

OCR on PDFs requires two main steps: converting PDF pages to images, then running Tesseract on those images. Here's the complete pipeline using pdftoppm and Tesseract. First, convert the PDF pages to high-resolution images: `pdftoppm -r 300 scanned.pdf /tmp/page` This creates /tmp/page-1.ppm, /tmp/page-2.ppm, etc. at 300 DPI — a high enough resolution for accurate OCR. Then run Tesseract on each page to extract text: ```bash for img in /tmp/page-*.ppm; do tesseract "$img" stdout >> extracted_text.txt done ``` Alternatively, create a searchable PDF directly from a single image: `tesseract page-1.ppm output_page pdf` This creates output_page.pdf which contains both the original scanned image and an invisible text layer — making it searchable while preserving the visual appearance. For a complete multi-page searchable PDF, the workflow uses Tesseract's built-in PDF output and then merges the pages: ```bash for img in /tmp/page-*.ppm; do name=$(basename "$img" .ppm) tesseract "$img" "/tmp/$name" pdf done pdftk /tmp/page-*.pdf cat output searchable.pdf ```

Tesseract's accuracy depends heavily on image quality and configuration. Here are the most important techniques for improving OCR results on Linux. Use 300 DPI minimum when converting PDF pages to images with pdftoppm. Lower resolutions degrade OCR accuracy significantly — at 72 DPI, Tesseract struggles with many fonts. At 300 DPI, accuracy on clean documents is typically 98%+. Specify the correct language with the `-l` flag: `tesseract image.ppm output -l fra+eng` This uses French as the primary language with English as secondary — useful for bilingual documents. Use the OEM (OCR Engine Mode) flag for better accuracy. Mode 1 uses LSTM (neural network), which is Tesseract's most accurate engine: `tesseract image.ppm output -l eng --oem 1` Preprocess images for better results using ImageMagick before OCR: ```bash convert page-1.ppm -colorspace Gray -threshold 50% preprocessed.ppm tesseract preprocessed.ppm output ``` Grayscale conversion and thresholding can significantly improve accuracy on low-contrast or colored backgrounds.

If the manual pipeline feels complex, OCRmyPDF is a high-level tool that wraps Tesseract and handles the entire PDF OCR workflow in a single command. It's the most user-friendly OCR solution for Linux PDF files. Install OCRmyPDF: `sudo apt install ocrmypdf` Basic usage: `ocrmypdf input.pdf output.pdf` This adds a text layer to the PDF in place, making it searchable while preserving the original scanned images. The output is a standard PDF that opens in any viewer. OCRmyPDF handles multi-page PDFs automatically, includes image preprocessing (deskew, cleanup), and even repairs some malformed PDFs. It uses Tesseract internally but abstracts away all the intermediate steps. For more control: `ocrmypdf -l fra+eng --deskew --clean input.pdf output.pdf` The `--deskew` flag automatically corrects slightly tilted scans, and `--clean` removes background noise — both significantly improve OCR accuracy on real-world scanned documents.