Absolute power to automate OCR.
$ npm install -g autocracy
Alternatively, don't install it and just prepend the below commands with npx.
Completions for Zsh will also be installed if a directory exists:
$ mkdir -p /usr/local/share/zsh/site-functions
$ chown -R $(whoami) /usr/local/share/zsh/site-functions
You will also need to install tools that Autocracy relies on to operate, Tesseract, MuPDF, and QPDF. On a Mac with Homebrew these can be installed with brew install tesseract mupdf qpdf. With Apt you will need to run apt install tesseract-ocr mupdf-tools qpdf.
If not using Homebrew check your Tesseract installation includes the fast training data for your desired languages, which can otherwise be downloaded from here.
To output text files:
$ autocracy get-text <origin> <destination>
To output new PDF files with embedded text:
$ autocracy make-searchable <origin> <destination>
In either case, the origin should be a directory of PDF files. The destination should be the name of a directory to be created for the results.
By default, Autocracy will first attempt to extract any tagged-text from within the PDF files. If tagged-text is found, it is used instead of (much slower) OCR. To disable this use the --force-ocr flag. The --preprocess flag will do some processing to attempt to improve OCR quality. The language expected in the documents defaults to English, but can be specified by passing the --language flag one of the language codes from this page.
A directory named .autocracy-cache will be created to contain intermediate files. These will be used on subsequent invocations of Autocracy. You will want to delete this directory after you finish.