I’ve been looking for an OCR engine and hoped that using LLMs would improve their output.

Looks great! I’ll give it a go.