EasyOCR — Ready-to-Use OCR for 80+ Languages

EasyOCR

EasyOCR provides production-ready optical character recognition for over 80 languages with a minimal API. Built on CRAFT text detection and a ResNet-LSTM-CTC recognition backbone, it handles complex scripts, mixed-language text, and degraded images robustly.

Key Features

• 80+ language support out of the box (Latin, CJK, Arabic, Cyrillic, Devanagari, …)
• GPU and CPU inference with automatic device selection
• Single-line API: readtext(image) returns bounding boxes + text + confidence
• Paragraph mode for flowing text reconstruction
• Allowlist/blocklist for domain-specific vocabularies (ID numbers, license plates)
• CLI: easyocr -l en -f image.jpg --detail 1

Quick Start

import easyocr

reader = easyocr.Reader(["en", "fr"])  # load once
results = reader.readtext("invoice.png")

for bbox, text, confidence in results:
    print(f"{text!r:40s} {confidence:.2f}")

Install via ai-supply

npx ai-supply add easyocr-multilingual-ocr

Curated mirror of the open-source EasyOCR (Apache-2.0). Get it from the source.