---
id: rapidocr-onnxruntime
version: "1.4.4"
license: Apache-2.0
license_treatment: permissive
maintenance: active
---

# rapidocr-onnxruntime — A cross platform OCR Library based on OnnxRuntime.

License: permissive · Maintenance: active · Downloads: 2.5M/mo

## What it is and what it does

RapidOCR is an open-source OCR library that converts images to text using ONNX-format models derived from PaddleOCR. It wraps nine runtime dependencies—opencv-python, numpy, Pillow, onnxruntime, and others—to provide text detection and recognition in a single Python interface. The package is designed for offline deployment with minimal resource consumption and cross-platform compatibility.

You instantiate a RapidOCR engine and call it on image paths or URLs; it returns structured results containing detected text boxes and their content. The library supports Chinese and English by default, with other languages available via model substitution. It is actively maintained, widely used in downstream projects (langchain, Docling, and others), and carries no known security vulnerabilities.

Use it for:

- Extract text from scanned documents or screenshots for archival or processing pipelines.
- Build document understanding workflows that combine OCR with language models for semantic analysis.
- Automate data entry by recognizing text in forms, receipts, or invoices without cloud API calls.
- Integrate OCR into desktop or embedded applications requiring offline text recognition.
- Detect and extract multilingual text from images in real-time applications.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

Performs optical character recognition (OCR) on images to extract text, supporting Chinese, English, and other languages via ONNX Runtime inference.

Yes. The package is actively maintained, has low install friction, carries no known vulnerabilities, and is widely adopted in production systems. Apache 2.0 licensing is permissive. Install it if you need offline OCR with reasonable speed and accuracy; verify model download behavior and performance on your target hardware before committing to production use.

## Install

pip install rapidocr-onnxruntime
uv add rapidocr-onnxruntime
poetry add rapidocr-onnxruntime

## Installing rapidocr-onnxruntime

Before you install: Low friction: pure Python wheel with no compiled dependencies beyond onnxruntime. Active maintenance with recent commits and 7492 repository stars. Supports Python 3.6 through 3.12.

License in practice: Apache 2.0 permissive license. The OCR model copyright is held by Baidu; engineering scripts are owned by the repository. You may use, modify, and distribute freely with attribution.

Quickstart:

pip install rapidocr-onnxruntime onnxruntime

from rapidocr import RapidOCR

engine = RapidOCR()
result = engine("path/to/image.jpg")
print(result)

onnxruntime must be installed separately; models are downloaded on first use and require internet access or manual placement.

Verify before relying:

- Whether model download/caching behavior and disk space requirements are documented
- Performance characteristics (inference speed, memory usage) on typical hardware
- Accuracy metrics for supported languages beyond Chinese and English

## Package facts

- License: Apache-2.0 (permissive)
- Python support: capped_below_current
- Install friction: low
- Maintenance: active
- Downloads: 2.5M/month (top 5,000 on PyPI)
- Known vulnerabilities: none known

## Tags

ocr text extraction, optical character recognition, image to text, chinese english ocr, offline ocr deployment, onnx ocr engine, text detection recognition, ocr, computer-vision, text-extraction

[View on SkillFed](https://skillfed.io/packages/rapidocr-onnxruntime) · [View on PyPI](https://pypi.org/project/rapidocr-onnxruntime/)