Extracts all UTF-8 strings from YouTube's closed caption ("timedtext") JSON object to generate a rough transcript.

extract_youtube_closed_captions.rb

require 'json'

# Extracts all UTF-8 strings from YouTube's closed caption (`timedtext`) JSON 
# object to generate a rough transcript.
# 
# Instructions for acquiring the JSON: 
# 
# 1. Open the video in a new Firefox tab. Don't click "CC" yet.
# 
# 2. Open Firefox Developer Tools (right-click somewhere and select "Inspect 
#    Element") and open the "Network" tab. (Shortcut: Ctrl+Shift+E)
# 
# 4. Click "CC" to turn on closed captions.
# 
# 5. Right click on the "?timedtext(somethingsometing)" object that appears 
#    in the list.
# 
# 6. Select "Copy response".
# 
# 7. Paste the JSON into your text editor / console / whatever.
# 
def extract_closed_captions(json_string, output_fname="video_transcript.txt")
  json  = JSON::parse(json_string, symbolize_names: true)
  words = Array.new
  json[:events].each do |event|
    next unless event[:segs].is_a?(Array)
    event[:segs].each do |segment|
      next unless segment.is_a?(Hash)
      words.push(segment[:utf8])
    end
  end
  transcript = words.join
  File.write(output_fname, transcript) if output_fname
  return transcript
end