LMKitOcr
Native: built for throughput
High-throughput recognition tuned for business documents: invoices, contracts, reports, forms. Page layout analysis with intelligent reading-order reconstruction, automatic language and orientation detection, deskew, despeckle, and smart binarization built in. 35 language packs download on demand; concurrency is gated to physical cores; every phase raises an event (LanguageDetected, OrientationDetected, OcrStarting, OcrCompleted). CPU-only, no GPU required.