Recent versions of iOS and macOS offer quite good OCR for images, which suggested that these systems must now ship with an OCR library. It turns out they do, so I've made a tiny command line OCR utility that takes image paths as arguments and emits text to stdout.

This replaces a more elaborate system I've been maintaining for years with something very minimal. 🥳

Clone the repo, cd into the top level, run:

swift build -c release

The resulting binary should appear in ./.build/release/tinyocr. Copy it to a directory on your path and you should be able to:

tinyocr <image>