ocrs

Why we included this project

ocrs is a Rust OCR engine that skips most of the preprocessing older tools demand. It leans on neural models trained on open datasets, so scanned documents, photos with text, and screenshots usually come through without the thresholding and scaling Tesseract often needs. The CLI installs with a single cargo command, and the library runs its ONNX models locally through RTen, which matters if you care about on-device processing. It compiles to WebAssembly too, so you can drop it into browser-based tools. Just know it's early preview and Latin-only for now, so don't expect commercial-grade accuracy yet.