TL;DR
PaddleOCR is a lightweight OCR toolkit that converts PDF or image documents into structured JSON/Markdown data, supporting over 100 languages.
Key features
Extracts text from PDF and image documents and outputs structured data (JSON/Markdown)
Supports over 100 languages, runs on various hardware (CPU, GPU, XPU, NPU) with lightweight models
Easy to install and use via Python package (paddleocr)
When to use it
When you need to extract text from PDF or image documents (contracts, reports, scans, etc.) for input into LLMs or data pipelines
When you need OCR for multilingual documents
When you want a lightweight OCR solution