easyocr.js

easyocr.js is the first complete JavaScript/TypeScript port of EasyOCR. It reimplements the whole pipeline (image preprocessing, CRAFT text detection, CRNN recognition, CTC decoding) and runs it on ONNX Runtime, so it needs no Python and spawns no Python processes.

Compared with Python-based EasyOCR wrappers

This comparison is specifically with JavaScript packages that call EasyOCR’s Python implementation. It does not describe JavaScript OCR libraries generally: for example, Tesseract.js runs in browsers and Node.js using WebAssembly.

	Python-backed EasyOCR wrapper	easyocr.js
Implementation	Calls EasyOCR’s Python implementation	Native JavaScript/TypeScript port
Python at runtime	Required	Not required
Inference	EasyOCR’s Python runtime	ONNX Runtime in Node.js or the browser

Features

• Same accuracy as Python: Outputs match the Python reference implementation, checked by validating and tracing each pipeline stage.
• Same models: Uses the original CRAFT detector and recognition models, with 80+ languages including English, Chinese, Japanese, and Korean.
• Node.js and browser: Node uses sharp and onnxruntime-node; the browser uses onnxruntime-web and Canvas. The core has no runtime-specific dependencies.
• CPU/GPU acceleration: Through ONNX Runtime, with no platform-specific bindings.
• No Python: No interpreter, virtual environment, or pip packages to install.
• Runs client-side: In the browser, documents never leave the machine.
• TypeScript: Fully typed.

Quick Start

# Node.js
npm install @qduc/easyocr-node @qduc/easyocr-core

# Browser
npm install @qduc/easyocr-web @qduc/easyocr-core

Tech Stack

• TypeScript
• ONNX Runtime
• Sharp (Node.js)
• Canvas API (Browser)

Learn more in the GitHub repository.