google_sr.py

import speech_recognition as sr
import argparse
import os
from tqdm import tqdm

args = argparse.ArgumentParser()
args.add_argument('--audio-input-dir', '-i', type=str, dest='audio_input_dir', required=True)

args.add_argument('--txt-output-path', '-o', type=str, dest='txt_output_dir', required=False, default='')
args.add_argument('--lang', '-l', type=str, dest='lang', required=False, default='ko-KR')
args = args.parse_args()

if not os.path.exists(args.audio_input_dir):
    raise Exception('audio_input_dir is not exist')

if args.txt_output_dir == '':
    args.txt_output_dir = os.path.join(args.audio_input_dir, 'manifest.txt')

r = sr.Recognizer()
txt = open(args.txt_output_dir, 'w', encoding='utf-8')

for file in tqdm([i for i in os.listdir(args.audio_input_dir) if i.endswith('.wav')]):
    audio_input = os.path.join(args.audio_input_dir, file)
    with sr.AudioFile(audio_input) as source:
        audio = r.record(source)
    try:
        response = r.recognize_google(audio, language=args.lang)
        txt.write(f"{file}|{response}\n")
    except sr.UnknownValueError:
        print('UnknownValueError')
    except sr.RequestError as e:
        print('RequestError: {0}'.format(e))

txt.close()