Extract closed caption text from a YouTube video

subtitle-clean.py

#!/usr/bin/env python3
# Don't call directly, called by `subtitles.sh`

"""Given an srt file, clean it up as much as possible to look like prose."""

import re
import sys

bad_words = ['-->', 'WEBVTT', 'Language: en', 'Kind: captions']

def cleanline(raw_line):
    cleanr = re.compile('<.*?>')
    cleantext = re.sub(cleanr, '', raw_line)
    return cleantext

vid_id = sys.argv[1]
with open(vid_id) as oldfile, open('newfile.txt', 'w') as newfile:
    for line in oldfile:
        if not any(bad_word in line for bad_word in bad_words):
            line = cleanline(line)
            newfile.write(line)


with open('newfile.txt') as result:
    uniqlines = set(result.readlines())
    with open(f'sub_{vid_id}.txt', 'w') as rmdup:
        mylst = map(lambda each: each.strip("&gt;&gt;"), uniqlines)
        print(mylst)
        rmdup.writelines(set(mylst))

subtitles.sh

#!/usr/bin/env bash
# Usage:
#   ./subtitles.sh tc4ROCJYbm0
# 
# Invokes `subtitle-clean.py`
# NOTE: Downloads youtube-dl Python script if this is our first time
cd /tmp
if [ ! -f /tmp/youtube-dl ] 
then
    echo "Downloading youtube-dl script..."
    curl -L -O https://github.com/ytdl-org/youtube-dl/releases/latest/download/youtube-dl
fi

if [ ! -x /tmp/youtube-dl ]
then
    chmod +x /tmp/youtube-dl
fi

echo "Downloading subtitles..."
VIDEO=$1
./youtube-dl --skip-download  --convert-subs srt --write-auto-sub --id --sub-lang en $VIDEO

if [ ! -x ~/subtitle-clean.py ]
then
    chmod +x ~/subtitle-clean.py
fi

~/subtitle-clean.py ${VIDEO}.en.vtt
cat sub_${VIDEO}.en.vtt.txt