A wrapper around mistral's OCR I built to auto-magically transcribe photos of my hand written notes into markdown.
Run monitor_ocr.sh in the background. It will monitor the specified directory (in this case /test). Whenever an image or PDF is copied to the directory, it will extract the writing for them photos and any diagrams / drawings / photos, and dump them in the specified output directory. (In this case /outputs)
This is hitting an external server and they are training on your data. Keep that in mind before using it on anything sensitive. Also all LLMs have issues with jank, don't use for anything mission critical.