OCR Language Packs and Setup

This document provides instructions on how to add additional language packs for the OCR tab in Stirling PDF, both inside and outside of Docker.

Looking to use OCR?

This page covers server setup. For how the OCR tool works and its options, see OCR (Functionality).

How does the OCR Work#

Stirling PDF uses Tesseract for its text recognition. All credit goes to them for this awesome work! Note that OCR recognizes text only - it does not perform table-structure or formula recognition.

Requires a server backend

OCR runs on a server-side backend with Tesseract installed. The desktop app cannot OCR in local-only mode - connect it to Stirling Cloud or a self-hosted server that has OCR available. To OCR non-English documents, install the matching language pack as described below.

Language Packs#

Tesseract OCR supports a variety of languages. You can find additional language packs in the Tesseract GitHub repositories:

tessdata_fast: These language packs are smaller and faster to load but may provide lower recognition accuracy.
tessdata: These language packs are larger and provide better recognition accuracy, but may take longer to load.

Depending on your requirements, you can choose the appropriate language pack for your use case. By default, Stirling PDF uses tessdata_fast for English, but this can be replaced.

Installing Language Packs manually#

Download the desired language pack(s) by selecting the .traineddata file(s) for the language(s) you need.
Place the .traineddata files in the Tesseract tessdata directory: /usr/share/tessdata (or equivalent)

DO NOT REMOVE EXISTING eng.traineddata, IT'S REQUIRED.

Docker Setup#

If you are using Docker, you need to expose the Tesseract tessdata directory as a volume in order to use the additional language packs.

Modify your docker-compose.yml file to include the following volume configuration:

yaml

services:
  your_service_name:
    image: your_docker_image_name
    volumes:
      - /location/of/trainingData:/usr/share/tessdata

Add the following to your existing Docker run command:

bash

-v /location/of/trainingData:/usr/share/tessdata

Non-Docker Setup#

For Debian-based systems, use the following commands to manage Tesseract languages:

bash

sudo apt update &&\
# All languages
# sudo apt install -y 'tesseract-ocr-*'

# Find available languages:
apt search tesseract-ocr-

# View installed languages:
dpkg-query -W tesseract-ocr- | sed 's/tesseract-ocr-//g'

For Fedora systems, use the following commands:

bash

# All languages
# sudo dnf install -y tesseract-langpack-*

# Find available languages:
dnf search -C tesseract-langpack-

# View installed languages:
rpm -qa | grep tesseract-langpack | sed 's/tesseract-langpack-//g'

Follow these steps to set up Tesseract languages on Windows:

Download desired .traineddata files from tessdata or tessdata_fast
Place them in the tessdata folder within your Tesseract installation directory:
```
C:\Program Files\Tesseract-OCR\tessdata
```
Verify the installation by running:
powershell
```
tesseract --list-langs
```

Edit your /configs/settings.yml and update the system.tessdataDir:

yaml

system:
  tessdataDir: C:/Program Files/Tesseract-OCR/tessdata # path to Tessdata files

Edit this page on GitHub