mirrors/tesseract.js

npm jsDelivr hits (npm)

Tesseract.js is a javascript library that gets words in almost any language out of images. (Demo)

Image Recognition

Video Real-time Recognition

Tesseract.js works in the browser using webpack, esm, or plain script tags with a CDN and on the server with Node.js. After you install it, using it is as simple as:

import { createWorker } from 'tesseract.js';

(async () => {
  const worker = await createWorker('eng');
  const ret = await worker.recognize('https://tesseract.projectnaptha.com/img/eng_bw.png');
  console.log(ret.data.text);
  await worker.terminate();
})();

When recognizing multiple images, users should create a worker once, run worker.recognize for each image, and then run worker.terminate() once at the end (rather than running the above snippet for every image).

Installation

Tesseract.js works with a <script> tag via local copy or CDN, with webpack via npm and on Node.js with npm/yarn.

CDN

<!-- v5 -->
<script src='https://cdn.jsdelivr.net/npm/tesseract.js@5/dist/tesseract.min.js'></script>

After including the script the Tesseract variable will be globally available and a worker can be created using Tesseract.createWorker.

Alternatively, an ESM build (used with import syntax) can be found at https://cdn.jsdelivr.net/npm/tesseract.js@5/dist/tesseract.esm.min.js.

Node.js

Tesseract.js v7 requires Node.js v16 or newer. (Tesseract.js v6 requires Node.js v14 or newer.)

# For latest version
npm install tesseract.js
yarn add tesseract.js

# For old versions
npm install tesseract.js@3.0.3
yarn add tesseract.js@3.0.3

Project Scope

Tesseract.js aims to bring the Tesseract OCR engine (a separate project) to the browser and Node.js, and works by wrapping a WebAssembly port of Tesseract. This project does not modify core Tesseract features. Most notably, Tesseract.js does not support PDF files and does not modify the Tesseract recognition model to improve accuracy.

If your project requires features outside of this scope, consider the Scribe.js library. Scribe.js is an alternative library created to accommodate common feature requests that are outside of the scope of this repo. Scribe.js includes improvements to the Tesseract recognition model and supports extracting text from PDF documents, among other features. For more information see Scribe.js vs. Tesseract.js.

Documentation

Community Projects and Examples

The following are examples and projects built by the community using Tesseract.js. Officially supported examples are found in the examples directory.

Projects
- Scribe OCR: web application for scanning documents (images and PDFs)
  - Site at scribeocr.com, repo at github.com/scribeocr/scribeocr
- Chrome Extension (with Manifest V3): https://github.com/Tshetrim/Image-To-Text-OCR-extension-for-ChatGPT
Examples
- Converting PDF to text: https://github.com/racosa/pdf2text-ocr
- Use blocks output to generate granular data [word/symbol level]: https://github.com/Kishlay-notabot/tesseract-bbox-examples
- Electron: https://github.com/Balearica/tesseract.js-electron
- Typescript: https://github.com/Balearica/tesseract.js-typescript

If you have a project or example repo that uses Tesseract.js, feel free to add it to this list using a pull request. Examples submitted should be well documented such that new users can run them; projects should be functional and actively maintained.

Major changes in v6

Version 6 changes are documented in this issue. Highlights are below.

Fixed memory leak in previous versions
Overall reductions in runtime and memory usage
Breaking changes:
- All outputs formats other than text are disabled by default.
  - To re-enable the hocr output (for example), set the following: worker.recognize(image, {}, { hocr: true })
- Minor changes to the structure of the JavaScript object (blocks) output
- See this issue for full list

Major changes in v5

Version 5 changes are documented in this issue. Highlights are below.

Significantly smaller files by default (54% smaller for English, 73% smaller for Chinese)
- This results in a ~50% reduction in runtime for first-time users (who do not have the files cached yet)
Significantly lower memory usage
Breaking changes:
- createWorker arguments changed
  - Setting non-default language and OEM now happens in createWorker
    - E.g. createWorker("chi_sim", 1)
- worker.initialize and worker.loadLanguage functions should be deleted from code
- See this issue for full list

Upgrading from v2 to v5? See this guide.

Major changes in v4

Version 4 includes many new features and bug fixes–see this issue for a full list. Several highlights are below.

Added rotation preprocessing options (including auto-rotate) for significantly better accuracy
Processed images (rotated, grayscale, binary) can now be retrieved
Improved support for parallel processing (schedulers)
Breaking changes:
- createWorker is now async
- getPDF function replaced by pdf recognize option

Contributing

Development

To run a development copy of Tesseract.js do the following:

# First we clone the repository
git clone https://github.com/naptha/tesseract.js.git
cd tesseract.js

# Then we install the dependencies
npm install

# And finally we start the development server
npm start

The development server will be available at http://localhost:3000/examples/browser/basic-efficient.html in your favorite browser. It will automatically rebuild tesseract.min.js and worker.min.js when you change files in the src folder.