ocrmypdf/OCRmyPDF v17.14.1

on GitHub

Fixes

• Fixed the RapidOCR engine offering no way to choose where its models are
  kept. They could only live in RapidOCR's package directory, which may be
  read-only, and models downloaded into a container's directory were lost
  when it exited. The new option --rapidocr-model-dir DIR (API:
  rapidocr_model_dir=; environment: OCRMYPDF_RAPIDOCR_MODEL_DIR) keeps
  all of RapidOCR's models, and the page orientation model, in the given
  directory, downloading any that are missing into it. python -m ocrmypdf.builtin_plugins.rapidocr_ocr download --model-dir DIR fills it in
  advance.