I’m looking for an on-premises OCR solution that can be implemented in C++ for a high-accuracy document scanning task. The documents will mainly be photos taken with smartphones, including passports, bills, and receipts, which are likely to be skewed and they will mostly contain printed data. I’ve explored several options but found them lacking:

Tesseract OCR, OCR.space, EasyOCR, PaddleOCR, CuneiForm, and GOCR had low accuracy.
Paid solutions like ABBYY FineReader, Google Cloud Vision OCR, and Microsoft Azure Computer Vision API are too expensive.
I need an OCR solution that meets the following criteria:

-On-premises
-Low cost or free
-Implementable in C++ or has a C++ implementation
-High accuracy

please recommend me any possible solution.