Transribe youtube video using vosk.

download.py

import re
import yt_dlp

def download_wav(url):
  ydl_opts={}

  with yt_dlp.YoutubeDL(ydl_opts) as ydl:
    info_dict = ydl.extract_info(url, download=False)

  video_title = info_dict['title']
  video_name = re.sub('[\\\\/*?:"<>|]', '', video_title)
  name = video_name
  ydl_opts = {
      'format': 'm4a/bestaudio/best',
      'noplaylist': True,
      'continue_dl': True,
      'outtmpl': f'./{name}.wav',
      'postprocessors': [{
          'key': 'FFmpegExtractAudio',
          'preferredcodec': 'wav',
          'preferredquality': '192',
      }],
      'geobypass':True,
      'ffmpeg_location':'/opt/homebrew/bin/ffmpeg'
  }

  with yt_dlp.YoutubeDL(ydl_opts) as ydl:
    error_code = ydl.download(url)
  print(error_code)

  file_name = "{}.wav".format(video_name)
  return file_name

main.py

import os
import json
import wave

from vosk import Model, KaldiRecognizer
from pydub import AudioSegment
from tqdm import tqdm
from download import download_wav

def saveListToFile(values, file_path):
    file1 = open(file_path,"w") 
    for val in values:
        file1.writelines("{}\n".format(val))    
    file1.close() 

CHANNELS=1

def voice_recognition(filename):
    model = Model(model_name="vosk-model-en-us-0.22")
    wf = wave.open(filename, "rb")
    frame_rate = wf.getframerate()

    rec = KaldiRecognizer(model, frame_rate)
    rec.SetWords(True)
    
    wav_audio = AudioSegment.from_wav(filename)
    wav_audio = wav_audio.set_channels(CHANNELS)
    wav_audio = wav_audio.set_frame_rate(frame_rate)
    
    step = 30 * 1000 # 5 seconds
    transcript = []
    for i in tqdm(range(0, len(wav_audio), step)):
        segment = wav_audio[i:i+step]
        rec.AcceptWaveform(segment.raw_data)
        result = rec.Result()
        text = json.loads(result)["text"]
        transcript.append(text)
    return transcript

url = "https://www.youtube.com/shorts/03tVbko1RaY"
file_name = download_wav(url)
result = voice_recognition(filename=file_name)
saveListToFile(result, file_name.replace("wav", "txt"))
os.remove(file_name)