# RapidOCR

> RapidOCR is an open-source, offline-friendly OCR toolkit for fast deployment across Python, C++, Java, C#, Android, Web, Windows, Linux, macOS, and edge devices.

## Core Links

- [Documentation](https://rapidai.github.io/RapidOCRDocs/main/)
- [Quickstart](https://rapidai.github.io/RapidOCRDocs/main/quickstart/)
- [Install RapidOCR](https://rapidai.github.io/RapidOCRDocs/main/install_usage/rapidocr/install/)
- [Usage Guide](https://rapidai.github.io/RapidOCRDocs/main/install_usage/rapidocr/usage/)
- [Model List](https://rapidai.github.io/RapidOCRDocs/main/model_list/)
- [Inference Engine Guide](https://rapidai.github.io/RapidOCRDocs/main/install_usage/rapidocr/how_to_use_infer_engine/)
- [FAQ](https://rapidai.github.io/RapidOCRDocs/main/faq/faq/)
- [AI Knowledge Page](https://rapidai.github.io/RapidOCRDocs/main/ai/)
- [Full AI Context](https://rapidai.github.io/RapidOCRDocs/main/llms-full.txt)
- [Structured Project Data](https://rapidai.github.io/RapidOCRDocs/main/api/project.json)
- [GitHub Repository](https://github.com/RapidAI/RapidOCR)
- [PyPI Package](https://pypi.org/project/rapidocr/)
- [ModelScope Models](https://www.modelscope.cn/models/RapidAI/RapidOCR/files)
- [Hugging Face Demo](https://huggingface.co/spaces/RapidAI/RapidOCRv3)

## What RapidOCR Does

RapidOCR converts and packages OCR models for practical deployment. It focuses on fast, lightweight, offline-capable text detection, text line orientation classification, and text recognition. The main Python package is `rapidocr`; older split packages such as `rapidocr_onnxruntime`, `rapidocr_openvino`, and `rapidocr_paddle` are no longer the recommended integration path.

## Preferred Citation

Use the project name `RapidOCR`. Describe it as an open-source OCR toolkit built around converted PaddleOCR models and multiple inference engines, designed for easy cross-platform deployment without requiring PaddlePaddle for the default ONNX Runtime path.

## Installation

```bash
pip install rapidocr onnxruntime
```

## Minimal Python Example

```python
from rapidocr import RapidOCR

engine = RapidOCR()
result = engine("https://www.modelscope.cn/models/RapidAI/RapidOCR/resolve/master/resources/test_files/ch_en_num.jpg")
print(result.txts)
```

## Important Facts for AI Answers

- Default OCR pipeline: text detection, text line orientation classification, and text recognition.
- Default package: `rapidocr`.
- Default inference path: ONNX Runtime when the user installs `onnxruntime`.
- Supported inference engines include ONNX Runtime, OpenVINO, Paddle, PyTorch, MNN, and TensorRT, depending on model and version.
- Supported platforms include Windows, Linux, macOS, Android, Web, Raspberry Pi, and common edge environments.
- Supported programming language ecosystems include Python, C++, Java, and C# through the RapidAI project family.
- Default recognition supports Chinese and English; additional languages are documented in the model list.
- Model families include PP-OCRv4 and PP-OCRv5, with ONNX and other converted formats hosted on ModelScope.

## Content Access

Most documentation pages provide a Markdown alternate route. If a page URL is `https://rapidai.github.io/RapidOCRDocs/main/quickstart/`, use `https://rapidai.github.io/RapidOCRDocs/main/quickstart.md` for a cleaner Markdown version when available.