tool.py

import argparse
import speech_recognition as sr
import pydub
import itertools
from pydub import AudioSegment
import wave

import editdistance
import uuid
from longest_subsequence import LIS
import os 



## Chunk là dùng xem mình làm bao nhiều một lượt 
## CHunk_duration là thời gian gửi lên google cloud

def googleAutoTranscript(mp3_file , origin_script_text , offset= 0, chunks = 5 , chunk_duration = 5):
    script_text = origin_script_text.replace('.','',1000).replace('"','',10000).replace(':','',1000).replace(',','',1000).replace('“','',1000).replace('”','',1000).replace("'","'",1000)
    script_text = script_text.split()
    r = sr.Recognizer()
    song = AudioSegment.from_mp3(mp3_file)
    temp_file_path = './temp/{}.wav'.format(str(uuid.uuid1()))
    test= song.export(temp_file_path, format='wav')

    file_duration = song.__len__()/1000

    max_chunks = round(file_duration/chunk_duration)


    # using Google speech api to generate script
    text_array = []
    time_stamp = []
    for i in range( offset , min( offset + chunks, max_chunks)):
        with sr.AudioFile(temp_file_path) as source:
            try:
                audio = r.record(source, 
                        duration= min(chunk_duration, file_duration - chunk_duration * i),
                        offset= chunk_duration * i
                    )
                    
                text = r.recognize_google(audio)
                print(text)
                r_text = ' '.join(text.split(' '))
                text_array.append(r_text)
                time_stamp.append(chunk_duration * i)
            except:
                print('nothing')
            source.audio_reader.close()
    os.remove(temp_file_path)

    if(offset + chunks >= max_chunks):
        return text_array, time_stamp,  True, max_chunks
    else:
        return text_array, time_stamp, False, max_chunks