Tesseract OCR

Open Source Optical Character Recognition

Tesseract OCR is an open-source tool to read text in pictures and convert it into editable digital text. Perfect for scanning textbooks, documents, or notes into searchable formats. Useful for developers, students, and professionals who need accurate text extraction.

Download Now How to Use?

What is Tesseract OCR?

Tesseract is an Optical Character Recognition OCR software tool that extracts printed and, with training, some handwritten texts from pictures and PDFs and converts them into editable, machine-readable text. The original developer was Hewlett-Packard (HP) and was open sourced in 2005 has been community-maintained within the tesseract-ocr organization (Google sponsored development from 2006 to November 2018).

It supports more than 100 languages and is completely free, which is why it is available to everyone. The current stable version is Tesseract 5.x; Tesseract 4 introduced a modern LSTM (neural network) OCR engine that improved accuracy. It supports 100+ languages (the exact set depends on the installed trained data files).

Tesseract software is compatible with different programming languages and frameworks with wrappers like pytesseract for Python. It can be used directly via the command line or with an API.

OCR Made Simple with Tesseract

Key Features

Powerful, flexible, open-source OCR capabilities

🆓

Open Source

Free to use for personal and commercial work under Apache License 2.0.

🌐

100+ Languages

Supports over 100 languages and 37 scripts using trained data models.

🧠

LSTM Neural Engine

Modern LSTM-based recognition for higher accuracy on printed and handwritten text.

📄

Multiple Output Formats

Export to TXT, hOCR, TSV, PDF (searchable), or XML formats.

💻

Command Line Support

Process images quickly using CLI commands or automation scripts.

🔗

API Integrations

Use in Python (pytesseract), C++, Node.js, Java, and other frameworks.

🎓

Trainable Models

Train custom OCR datasets for new languages, fonts, and formats.

🔤

Unicode Support

Accurate handling of international scripts, accented characters, and symbols.

🗂️

Page Layout

Detects columns, borders, images, and paragraphs for better document structure.

Download Tesseract OCR

Grab the latest stable release and start extracting text in seconds.

⬇️ Latest Version: 5.5.1

🖥️ Download .exe (Windows) 📦 Download .zip 🧰 Download .tar.gz

How to Download and Install Tesseract OCR

Download and set up Tesseract for Windows, macOS, Linux and Python.

How to Install Tesseract OCR?

For Windows:

Download the .exe file of Tesseract OCR for Windows

Run the downloaded file and select your installer language.

Accept the terms and agreements.

In the components section, select your language (English).

Choose the installation directory (C:\Program Files\Tesseract-OCR) and copy it for configuration.

For Linux/Ubuntu:

Update system and install preconditions:

sudo yum update
sudo yum install git gcc-c++ make
sudo yum install openssl-devel
sudo yum install autoconf aclocal automake libtool
sudo yum install libjpeg-devel libpng-devel libtiff-devel zlib-devel
      

Then install Leptonica:

cd ~/libs
wget http://www.leptonica.com/source/leptonica-1.74.1.tar.gz
tar -zxvf leptonica-1.74.1.tar.gz
cd leptonica-1.74.1
./configure
make
sudo make install
      

Now install Tesseract OCR:

cd ~/libs
wget https://github.com/tesseract-ocr/tesseract/archive/3.05.00.tar.gz
tar -zxvf 3.05.00.tar.gz
cd tesseract-3.05-00
./autogen.sh
./configure
make
sudo make install
sudo ldconfig
      

Download language data files:

cd /usr/local/share/tessdata
sudo wget https://github.com/tesseract-ocr/tessdata/archive/3.04.00.zip
sudo unzip 3.04.00.zip
sudo mv tessdata-3.04.00/*.* .
sudo rm -r tessdata-3.04.00
      

Set environment variable:

sudo nano ~/.bash_profile

export TESSDATA_PREFIX=/usr/local/share/

source ~/.bash_profile

How To Launch Tesseract OCR?

Windows setup, Python usage, and Linux/macOS command-line examples.

Go to the System Environment Variables Settings and search for Environment Variables in Windows search.
Under the System Variables, look for the Path variable and Edit it.

Select New and paste C:\Program Files\Tesseract-OCR.
Now, in the same Environment Variables under System Variables, click New.

Set Variable name: TESSDATA_PREFIX and variable value as: C:\Program Files\Tesseract-OCR\ (the parent directory of tessdata), or you can use the --tessdata-dir option when running Tesseract commands.

Verify by opening a new Windows Command Prompt.
Run the command: tesseract -v.

If the installation was successful, the command will show the details of Tesseract OCR.

If you see an error like tesseract is not installed or it's not in your PATH, you need to point pytesseract to the executable location. Here's a complete example:

from PIL import Image
import pytesseract

# If tesseract is not in PATH, set the path to the executable:
pytesseract.pytesseract.tesseract_cmd = r"C:\Program Files\Tesseract-OCR\tesseract.exe"

# Load and process an image
img = Image.open("scan.png")
text = pytesseract.image_to_string(img, lang="eng", config="--psm 6")
print(text)
        

Run Python example for Tesseract (pytesseract)

Since Tesseract is a command line tool, you have to launch it from the terminal and there's no GUI application to open.

Easiest command: tesseract imagename outputbase to use English as the default language and generate a text file.

Common Examples:

Basic OCR (English) tesseract myimage.png output to create output.txt with extracted text.
Specify a language: tesseract myimage.png output -l eng. Here you have to replace eng with language codes like: spa for Spanish, fra for French, and hin for Hindi
For multiple languages: tesseract myimage.png output -l eng+deu Add -l LANG[+LANG] to use multiple languages together.
Create searchable PDFs: tesseract myimage.png output -l eng pdf to create a PDF and a separate searchable text layer.
Output to Terminal (stdout) tesseract myimage.png - -l eng The - outputs directly to the terminal instead of a file.
Check Installation tesseract --version

Tesseract OCR works the same way on macOS and Linux but the language data must be in the tessdata directory. The default output is a .txt file and that's it.

Video Overview

Features in Detail

A deeper look at what makes Tesseract OCR powerful

Supports 100+ Languages

Tesseract has robust Unicode (UTF-8) support and can recognize over 100 languages. Modern LSTM models (introduced in v4) expanded language coverage and quality, with many traineddata packs available for different scripts.

Supports Input Formats

Accepts common image formats such as PNG, JPEG, TIFF and more — making it compatible with screenshots, scanned pages, and images generated by deep-learning pipelines.

Supports Output Formats

Exports results to TXT, hOCR (HTML), TSV, searchable PDF, PAGE XML, and ALTO XML (availability depends on version and traineddata).

Engine Modes

Tesseract provides multiple engines: the legacy (pattern) engine and the modern LSTM neural engine. Use the --oem flag to choose: 0 (legacy), 1 (LSTM), 2 (both), or 3 (default/auto).

Page Layout

Detects document structure — columns, borders, images and paragraphs — and distinguishes monospace/code from proportional text for better extraction fidelity.

Trainable

Train or fine-tune LSTM models for new languages, fonts or specialized datasets. Training yields best results with high-quality ground-truth data and careful preprocessing.

API / Wrapper Access

Use Tesseract via its native C/C++ API or through language wrappers such as pytesseract (Python), tess4j (Java), and node-tesseract (Node.js).

Dual Engines (Summary)

Legacy and LSTM engines coexist to provide flexibility across use-cases. Choose the engine mode that fits your input quality and performance needs. Visual and theme styles in this section match the rest of the site.

Why Should You Use OCR?

OCR makes working with text faster, smarter, and more efficient in daily life and business workflows.

Faster than Manual Typing

OCR converts printed or handwritten text into editable digital text instantly. No need to retype entire pages.

Searchable Documents

Once converted, documents become searchable, indexable and organized for quick information retrieval.

Perfect for Study & Notes

Scan textbooks, notes, worksheets, and convert them to editable text for better studying and referencing.

Office Productivity

Convert scanned business papers, invoices, bills, receipts, ID cards, and official forms into usable digital text.

Supports Multiple Languages

OCR tools like Tesseract support 100+ languages, including English, Hindi, Arabic, Chinese, and more.

Useful in Automation & AI Tasks

OCR is widely used in machine learning, document processing systems, RPA workflows, and data extraction bots.

How Tesseract OCR Works

From raw image to clean, searchable text in four clear steps

STEP 1

Preprocessing

Clean the image for best results: deskew, denoise, convert to grayscale/threshold, and boost contrast so text stands out.

STEP 2

Layout Analysis

Detect text regions and structure (blocks → paragraphs → lines → words) so multi-column pages read correctly.

STEP 3

LSTM Recognition

Run the neural LSTM engine to read full lines of text, using the selected language model(s) for higher accuracy.

STEP 4

Post-process & Output

Spell/heuristic fixes and export: plain TXT, TSV, hOCR, or searchable PDF ready to copy, edit, and search.

STEP 1

Preprocessing

Clean the image for best results: deskew, denoise, convert to grayscale/threshold, and boost contrast so text stands out.

STEP 2

Layout Analysis

Detect text regions and structure (blocks → paragraphs → lines → words) so multi-column pages read correctly.

STEP 3

LSTM Recognition

Run the neural LSTM engine to read full lines of text, using the selected language model(s) for higher accuracy.

STEP 4

Post-process & Output

Spell/heuristic fixes and export: plain TXT, TSV, hOCR, or searchable PDF ready to copy, edit, and search.

The input image is cleaned up to maximize clarity, a process known as image preprocessing. Tesseract converts the image to greyscale, removes noise, and corrects rotation to align the text horizontally. After this, it performs Layout Analysis to identify and organize text into components such as blocks, paragraphs and singular lines.
Tesseract then performs Character Recognition, which in the legacy model is somewhat like this:
The older model recognizes the character patterns, and it separates the image into two blobs. The first blob uses a static classifier and the second one uses an adaptive classifier to improve accuracy.
The new model uses a neural network LSTM which is faster and quite modern. It recognizes whole lines of text rather than single characters like the older version.
The last step is called post processing, where the text is formatted and saved. It provides a variety of output options such as plain texts, structured formats such as hOCR that hold positional data and searchable PDFs. For developers, you can access it via a dedicated API or programming wrappers for Python.

How to Use Tesseract OCR?

OCR a file to text:

tesseract input.png outputbase -l eng

Multiple languages:

tesseract input.png outputbase -l eng+spa

Produce searchable PDF:

tesseract input.png outputbase -l eng pdf

Page segmentation mode (--psm)
Choose engine (--oem)
Print to console using stdout

For Python

Install the wrapper:

pip install pytesseract pillow

Load image with PIL and pass to pytesseract with config options.

pytesseract.image_to_string(image, lang="eng")

How to boost the speed of Tesseract OCR?

Use Native API (tesserocr) instead of wrappers.
Preload the API (PyTessBaseAPI) to avoid reloading models.
Optimize images (deskew, denoise, scale).
Set tessedit_do_invert=0 if text is already dark on light.
Limit OMP_THREAD_LIMIT=1 on old builds.
Benchmark multiple runs; ignore the first run.
Compare equal functions (do not compare line-by-line loops).

What do users say about Tesseract OCR?

Real feedback from teams, tools, and developers

Using Tesseract OCR has been a fantastic experience for our normal text extraction needs. We literally chose it because it’s completely free which saved us big bucks. Its ability to handle over 100 languages is remarkable. For clear and printed documents, the accuracy is quite good. For our workplace, Python wrappers helped a lot. Knowing that Google handles the operation makes it more reliable.

Tristan Thomman

Co-founder, Koncile

Source

We found Tesseract OCR accuracy on clean, printed files was quite high, and that made us even happier because it’s free as well. Being open source, the ability to recognize 100 languages with the LSTM engine was amazing.

Lizzy Lozano

Staff Editor, UPDF

Source

For us, Tesseract is our go-to workhorse for printed texts. This proven free OCR engine that offers accuracy at its best. For a beginner, it can be a little complex but other than that, the high quality multi language text extraction tool is the best.

SourceForge Review

Community Feedback

Source

Tesseract is our default OCR engine for its simplicity and wide language support, and no proprietary hurdles. We love how easy it is to set up and use. It even detects clear handwriting but not that clearly, but heavily redacted legal filings.

Sanjin Ibrahimovic

Developer Experience Engineer, MuckRock

Source

Tesseract is just amazing because of its open source foundation for any developer starting with OCR. The easy integration gets me each time. Love it.

Docsumo

Product Team Comment

Source

We found Tesseract highly flexible and useful. The best part is it’s free and open source, and you can actually train it according to your needs.

Klippa

Engineering Blog

Source

Community & Contributions

⭐ Star Tesseract OCR

Support the project by starring it on GitHub. Community support helps keep development active and growing.

Star on GitHub

🐞 Report Issues

Found a bug, unexpected output, or recognition problem? Help improve accuracy by reporting issues.

Report an Issue

🔄 Contribute Code / Fork

Want to improve the engine or add features? Fork the repository and submit development contributions.

Fork the Project

⚙️ View Build & CI Workflows

Check automated build pipelines, test systems, and CI/CD workflows powering Tesseract development.

View GitHub Actions

⭐ Star Tesseract OCR

Support the project by starring it on GitHub. Community support helps keep development active and growing.

Star on GitHub

🐞 Report Issues

Found a bug, unexpected output, or recognition problem? Help improve accuracy by reporting issues.

Report an Issue

🔄 Contribute Code / Fork

Want to improve the engine or add features? Fork the repository and submit development contributions.

Fork the Project

⚙️ View Build & CI Workflows

Check automated build pipelines, test systems, and CI/CD workflows powering Tesseract development.

View GitHub Actions

FAQs

Common questions answered clearly

Tesseract OCR is a famous open source Optical Character Recognition software that converts texts in images and PDFs into machine readable texts. It uses pattern recognition and neural network (LSTM) technology to recognize text characters and structures.

Being free and open source, it is considered the best text detection tool that you can actually train. You can customize and train the Tesseract software according to your needs and can use up to 100 languages and 37 scripts.

No, Tesseract OCR is not owned by Google. Google sponsored the development from 2006 till November 2018, but it has always been open source (released under Apache License 2.0). It is now maintained and managed by the community within the Tesseract-OCR GitHub organization. Google no longer maintains it.

Tesseract OCR is considered better because first of all, it’s free. Secondly, Tesseract is faster by 0.7 seconds on average. Google Cloud Vision is paid and cannot be trained or customized. Tesseract also has a better page layout analysis as well.

Tesseract OCR is absolutely free and open sourced. It is released under the Apache 2.0 license, which means you can use, modify, train, and share it freely for both personal and professional use or business projects.

Originally, Tesseract was developed by Hewlett-Packard (HP) between 1984 and 1995 and was open sourced in 2005, but later in 2006, Google sponsored the development till November 2018. Since 2018, it has been community-maintained, with contributors including Stefan Weil and the Mannheim University Library.

Yes, modern versions of Tesseract OCR like 4.0 and later use a system based on Long Short-Term Memory which is LSTM neural networks for text recognition. LSTM is particularly effective at recognizing whole lines of text rather than single characters.

Yes, Tesseract OCR is safe because it is open-source and hosted on GitHub, which maintains code integrity through version control and community review. When you launch Tesseract on your own device, your data stays local and will not be sent to external sources for processing.

Final Words

Tesseract OCR is a tool that you must try if you want an accurate detection of texts in images. Being a free and open source software, it is quite ahead of its time. The kind of accuracy it delivers is something to talk about. The best part is that it’s quite easy to use, but it can be a little complex for beginners.

If you want a free and better than most text detection software, then Tesseract OCR stands as one of the best on the market. Just a little practice, and you’ll master it.