A Python implementation of https://gist.github.com/aonemd/7bb3c4760d9e47a9ce8e270198cb40a0

subtitle_extractor.py

import os
from PIL import Image  # pip install Pillow
import pytesseract     # pip install pytesseract

newly_renamed = 0
not_renamed   = 0

for subdir, dirs, files in os.walk('./memes'):
    for file in files:
        src_path = os.path.join(subdir, file)
        tmp_path = os.path.join('./tmp/', file)

        img = Image.open(src_path).convert('L')
        bw  = img.point(lambda x: 0 if x < 249 else 250, '1')
        bw.save(tmp_path)
        img = Image.open(tmp_path)

        print("************************************")
        print(src_path)

        new_text = pytesseract.image_to_string(img).strip()
        if new_text:
            newly_renamed += 1
        else:
            not_renamed += 1
            new_text = src_path

        print(new_text)

print("**********************************************")
print(f"{newly_renamed} files newly renamed vs. {not_renamed} still have the same old name.")