natocr¶

natocr (native ocr) is a small Python wrapper around the OCR engines that already ship with macOS and Windows: the Vision framework on macOS and Windows Runtime OCR on Windows.

These built-in engines are generally faster, more efficient, and more accurate than third-party alternatives like Tesseract. natocr makes reaching for them painless via one clean Python API instead of wrangling with Objective-C bridges or WinRT async plumbing.

Install¶

pip install natocr

Add the extras group for JPEG XL / XR / HD and DjVu decoding (see supported file formats):

pip install natocr[extras]

Quick start¶

from natocr import OCR

ocr = OCR()
pages = ocr.recognize("invoice.png")

print(pages[0].text)

Invoice #1042 Total $58.20 Thank you!

recognize() always returns a list - one OCRResult per page. Most images are a single page, so reach for pages[0]; see Multi-page documents for DjVu / TIFF / GIF / animated PNG / multi-image HEIC/HEIF.

Bounding Boxes¶

page = ocr.recognize("receipt.png")[0]

for element in page.elements:
    box = element.bounds.bounds
    print(f"{element.text!r} @ {box} conf={element.confidence}")

'Acme Coffee' @ (24.0, 18.0, 180.0, 32.0) conf=0.97
'Latte' @ (24.0, 70.0, 96.0, 28.0) conf=0.95
'$4.50' @ (220.0, 70.0, 80.0, 28.0) conf=0.88

Confidence Scores¶

On macOS you can view the confidence score of the total detections or per individual detection. This is not currently available on windows.

page = ocr.recognize("drivers-license.jpg")[0]

# avg confidence, or None if unavailable
print(f"Overall confidence: {page.confidence}")

Overall confidence: 0.93

Next steps¶

Usage - confidence scores, bounding boxes, languages, accepted input types, supported file formats, and how to run the tests.
API Reference - the full OCR / OCRResult / TextElement / BoundingBox reference, generated from the source.