PaddleOCR

// Overview

PaddleOCR is an open-source toolkit for optical character recognition (OCR) and document parsing. It detects where text appears in an image and reads it back as machine-readable text, with support for more than 100 languages. Beyond plain text, it can turn PDFs and scanned images into structured outputs like JSON and Markdown.

It is aimed at developers who need to extract text or document structure from images at scale: think pipelines that feed scanned documents into search, retrieval-augmented generation (RAG), or LLM agents. The project ships ready-to-use models such as the PP-OCRv6 recognition models and the PP-StructureV3 document pipeline, so you do not have to train anything to get started.

As a computer-vision tool, PaddleOCR covers both general scene text spotting (IDs, street views, books, industrial parts) and layout-aware document conversion. It runs on Linux, Windows, and macOS, and supports CPU and GPU backends.

// What it does

• Recognizes text in 100+ languages, with PP-OCRv6 covering 50 languages (Chinese, English, Japanese, and 46 Latin-script languages) in a single unified model
• Converts complex PDFs and images into structured Markdown or JSON using the PP-StructureV3 pipeline, including table and text coordinates
• Includes the PaddleOCR-VL document vision-language model for parsing tables, formulas, and other document elements
• Handles natural scene text such as IDs, street views, books, and industrial components
• Offers PP-OCRv6 in three size tiers (tiny, small, medium) for edge, mobile, and server deployment
• Runs on Linux, Windows, and macOS across CPU, GPU, and other AI accelerators

// Getting started

Install PaddleOCR and an inference engine, then run text detection and recognition on an image with a few lines of Python.

python -m pip install "paddleocr[all]"

python -m pip install paddlepaddle==3.2.0 -i https://www.paddlepaddle.org.cn/packages/stable/cpu/

from paddleocr import PaddleOCR

ocr = PaddleOCR(
    use_doc_orientation_classify=False,
    use_doc_unwarping=False,
    use_textline_orientation=False,
    engine="paddle"
)

result = ocr.predict("./image.png")
for res in result:
    res.print()

paddleocr ocr -i ./image.png \
    --use_doc_orientation_classify False \
    --use_doc_unwarping False \
    --use_textline_orientation False \
    --engine paddle

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Extracting text from scanned documents, receipts, or IDs to feed into search or data pipelines
• Converting PDFs and images into Markdown or JSON for retrieval-augmented generation (RAG) and LLM agents
• Reading multilingual text from photos and scans without switching models
• Recognizing text in real-world scenes such as street views, books, or industrial components

// How PaddleOCR compares

PaddleOCR alongside other open-source vision & understanding tools AI/TLDR tracks, ranked by GitHub stars.