https://blog.greg.technology/2024/01/02/how-do-you-ocr-on-a-mac.html

the greg technology blog [ ]

how do you ocr on a mac using the cli or just python for free

Jan 2, 2024

a kind reader reached out about all of my iphone alarms and asked how
specifically did I run OCR on my mac. I'm not one to gatekeep, so
here's how you can get yourself a nice local ocr 'service' that works
in the terminal/CLI/python and costs 0 dollars:

  * ((by the way sharing Shortcuts is really weird which is why I'm
    having you do this all manually...))
  * (also, most of the stuff below only needs to be done once - after
    that, actually ocr'ing is easy)
  * open the Shortcuts app in your Applications folder
  * click the "+" button to create a new shortcut
  * in the right sidebar, search for "extract text"

[step1]

  * you should see "Extract Text from Image" in the list of possible
    actions. drag it from the right sidebar into the main area on the
    left
  * in the "Extract text from ..." action that's now on the left, click
    on the pale "Image" blue-ish pill
  * pick "Shortcut Input"

[step2]

  * you'll see a big "Receive Any input from Nowhere" action appear
    above the "Extract text from Shortcut Input". it's all good,
    leave it as is
  * in the right sidebar, search for "copy". you should see "Copy to
    Clipboard". drag that action below the "Extract text from ..." one
      + dragging the "Copy to ..." action below the "Extract from ..." is
        really annoyingly hard. just keep dragging it really below
        and it will work at some point
  * your final setup should look like this:

[step3]

  * in the last action, make sure that after "Copy" it does say "Text
    from Image"!
  * you're like almost done!!
  * in the window title bar, type stuff to name the shortcut. because
    I'm from a generation, I recommend naming it "extract-text" or
    "ocr-text" or something simple with no spaces and all lowercase.
    sorry. press enter after naming the shortcut otherwise it won't
    remember the name (...??)

[step4]

  * ok you may be actually done here I think
  * try out the shortcut in your terminal:

shortcuts run ocr-text -i <A PATH TO SOME IMAGE>

  * then try command-V - do you see the resulting text? if yes,
    you're good to go!!!
  * from python, you can do the following:

import subprocess

file_path = '... some file path ...'
ocr_out = subprocess.check_output(
    f'shortcuts run ocr-text -i "{file_path}"', shell=True
)
print(ocr_out)

  * THAT'S IT! x

 
main site * github * e-mail * I was recently at the Recurse Center
working on a lotta things.
<<< Prev * The Recurse Webring * Next >>>