a python script that converts .mp4 to .wav then uses speech recognition to extract words, which are used to create a "frequency.txt" and some other useful distribution reports for each file as a .txt

frq.py

from speech_recognition import *
from os import walk, path
import glob
import wave
import nltk
import contextlib
from pydub import AudioSegment
from pydub.silence import split_on_silence
import subprocess

def split(filepath):
	sound = AudioSegment.from_file(filepath)
	chunks = split_on_silence(
	sound,
		min_silence_len = 500,
		silence_thresh = sound.dBFS - 16,
		keep_silence = 250, # optional
		)
	return chunks

def recognize(audio_file):
	r = Recognizer()
	inp = AudioFile(audio_file)
	with inp as source:
		audio = r.record(source)
		try:
			all_words = r.recognize_google(audio)
			return all_words
		except:
			return "No speech detected"

def recognize_chunks(chunk):
	# Create a silence chunk that's 0.5 seconds (or 500 ms) long for padding.
	silence_chunk = AudioSegment.silent(duration=500)

	# Add the padding chunk to beginning and end of the entire chunk.
	audio_chunk = silence_chunk + chunk + silence_chunk

	# Export the audio chunk with new bitrate.

	#print("Exporting tempchunk.wav")
	audio_chunk.export(
		".//tempchunk.wav",
		bitrate = "44100",
		format = "wav"
	)

	r = Recognizer()
	inp = AudioFile(".//tempchunk.wav")
	with inp as source:
		audio = r.record(source)
		try:
			all_words = r.recognize_google(audio)
			return all_words
		except:
			return "No speech detected"

def main():
	start = time.time()
	root_dir = "./"
	all_words = {}
	durations = {}
	frequency = []
	duration = 0
	wav_suffix = ".wav"

	for filename in glob.iglob(root_dir + "**/*.mp4", recursive=True):
		new_name = filename.replace('.mp4','')
		if not path.exists(new_name+wav_suffix):
			subprocess.call(['ffmpeg', '-i', filename, '-acodec',\
		 	'pcm_s16le',  '-ac', '1', '-ar', '16000', new_name+wav_suffix])

	for filename in glob.iglob(root_dir + "**/*.wav", recursive=True):
		if filename.endswith('.wav') and "tempchunk" not in filename:
			words = ""
			if path.exists(filename+".time.txt"):
				myfile = open(filename+".time.txt", "r")
				time_in_file = myfile.readline()
				print(time_in_file)
				duration = float(time_in_file)
			else:
				with contextlib.closing(wave.open(filename,'r')) as f:
					frames = f.getnframes()
					rate = f.getframerate()
					duration = frames / float(rate)
					with open(filename+".time.txt", "w") as o:
						dur = str(duration)
						o.write(dur + '\n')
			durations[filename] = duration
			print(filename, duration, "total time")
			print(filename, 'processing')
# we need to split files that are longer than like 2 minutes
# minimum chunk length
			target_length = 25 * 1000 # 25 seconds

			if path.exists(filename+".txt"):
				with open(filename+".txt", 'r') as fp:
					for count, line in enumerate(fp):
						pass
					print('Total Lines', count + 1)
				myfile = open(filename+".txt", "r")
				for line in myfile:
					words = words + line + ' '
			else:
				if duration > 120:
					print(f"total words so far in array from chunks: ", end ='' )
					chunks = split(filename)
					output_chunks = [chunks[0]]
					for chunk in chunks[1:]:
						if len(output_chunks[-1]) < target_length:
							output_chunks[-1] += chunk
						else:
							output_chunks.append(chunk)

					for chunk in output_chunks:
						tempwords = recognize_chunks(chunk)
						if tempwords != "No speech detected":
							words += tempwords
						#print(tempwords)
						prtial = str(len(words.split(' ')))
						print(prtial, end=" ")
				else:
					words = recognize(filename)

			all_words[filename] = words
			if not path.exists(filename+".txt"):
				with open(filename+".txt", "w") as o:
				    for i in words.split(' '):
				        o.write(i + ' \n')
			if not path.exists(filename+".distribution.txt"):
				with open(filename+".distribution.txt", "w") as o:
					# create a frequency distribution
					fdist = nltk.FreqDist(all_words[filename].split(' '))
					# print the top 150 most spoken words
					o.write("word frequency report for "+ filename.replace('wav','')+"\n\n")
					for w, count in fdist.most_common(150):
						if not w.isspace():
							report = w.ljust(20) + str(count) + " \n"
							o.write(report)
		else:
			print(filename, 'skipped')

	for key, value in all_words.items():
		if value != "No speech detected":
			mins = int( durations[key] / 60 )
			avg = value.count(" ")/mins
			secs = (( durations[key] / 60) - mins) * 60
			word_totals =   (value.count(" ") - value.count(" ") %2) /2
			frequency.append("{}: mins:{} secs:{} total words:{} words per min:{}".format(\
			key.replace(".wav","").ljust(90), str(int(mins)).ljust(2), str(int(secs)).ljust(2), str(int(word_totals)).ljust(2), avg))
		else:
			print("No speech detected in {}".format(key))

	with open("frequency.txt", "w") as o:
		for i in frequency:
			print(i+ '\n')
			o.write(i + '\n')

	if os.path.exists("tempchunk.wav"):
  		os.remove("tempchunk.wav")

	print(time.time() - start, "Total execution time")

if __name__ == '__main__':
	main()

Thrown together for fun to analyse the words per minute of the excellent videos by James Sharman
https://twitter.com/WeirdBoyJim
https://www.youtube.com/watch?v=3iHag4k4yEg&list=PLFhc0MFC8MiCDOh3cGFji3qQfXziB9yOw