def transcribe_file():
if request.method == 'POST':
res = {}
if 'file' not in request.files:
res['status'] = "error"
res['message'] = "audio file should be passed for the transcription"
return jsonify(res)
file = request.files['file']
filename = file.filename
_, file_extension = os.path.splitext(filename)
if file_extension.lower() not in ALLOWED_EXTENSIONS:
res['status'] = "error"
res['message'] = "{} is not supported format.".format(
file_extension)
return jsonify(res)
with NamedTemporaryFile(suffix=file_extension) as tmp_saved_audio_file:
file.save(tmp_saved_audio_file.name)
logging.info('Transcribing file...')
transcription, _ = run_transcribe(audio_path=tmp_saved_audio_file,
spect_parser=spect_parser,
model=model,
decoder=decoder,
device=device,
use_half=config.model.use_half)
logging.info('File transcribed')
res['status'] = "OK"
res['transcription'] = transcription[0][0]
return json.dumps(res, ensure_ascii=False).encode('utf8')
def transcribe_file():
if request.method == 'POST':
try:
res = {}
if 'file' not in request.files:
res['status'] = "error"
res['message'] = "audio file should be passed for the transcription"
return jsonify(res)
file = request.files['file']
filename = file.filename
_, file_extension = os.path.splitext(filename)
if file_extension.lower() not in ALLOWED_EXTENSIONS:
res['status'] = "error"
res['message'] = "{} is not supported format.".format(
file_extension)
return jsonify(res)
with NamedTemporaryFile(
suffix=file_extension) as tmp_saved_audio_file:
file.save(tmp_saved_audio_file.name)
logging.info('Transcribing file...')
transcription, _ = run_transcribe(
audio_path=tmp_saved_audio_file,
spect_parser=spect_parser,
model=model,
decoder=decoder,
device=device,
use_half=config.model.use_half)
logging.info('File transcribed')
res['status'] = "OK"
res['transcription'] = transcription[0][0]
return jsonify(res)
except Exception as e:
logging.error(e)
res['status'] = "error"
res['message'] = "The server encountered an internal error and was unable to complete your request."
return jsonify(res)
def transcribe_file():
if request.method == 'POST':
res = {}
res['total'] = 0
res['seconds'] = 0
t0 = time.time()
transTxt = ""
if 'file' not in request.files:
res['code'] = 403
res['data'] = "Missed audio files"
return jsonify(res)
try:
file = request.files['file']
filename = file.filename
_, file_extension = os.path.splitext(filename)
if file_extension.lower() not in ALLOWED_EXTENSIONS:
res['code'] = 403
res['data'] = "{} is not supported format.".format(
file_extension)
return jsonify(res)
with NamedTemporaryFile(prefix="product_",
suffix=file_extension,
dir='/work/dataset_product/wav',
delete=False) as temp_audio:
file.save(temp_audio.name)
path = temp_audio.name
if (file_extension.lower() == ".webm"):
# copyFile = temp_audio.name
# copyFile = copyFile + ".webm"
# strCopy = "cp {0} {1}".format(temp_audio.name,copyFile)
# os.system(strCopy)
# wavName = temp_audio.name
# wavName = wavName.replace("webm", "wav")
# strCv = "ffmpeg -i {0} -r 16000 -bits_per_raw_sample 16 -ac 1 {1}".format(temp_audio.name, wavName)
# os.system(strCv)
# path = wavName
#-------------------
#chuyen sang webm->mp3
src1 = temp_audio.name #.webm
dst1 = temp_audio.name #.webm
dst1 = dst1.replace("webm", "mp3") #.mp3
convertWebmToMp3(src1, dst1) #.wav
#chuyen mp3->wav 16000Hz
src2 = dst1
dst2 = dst1.replace("mp3", "wav")
convertMp3ToWav16(src2, dst2)
os.remove(dst1)
path = dst2
if (file_extension.lower() == ".mp3"):
#chuyen mp3->wav 16000Hz
src = temp_audio.name
dst = src.replace("mp3", "wav")
convertMp3ToWav16(src, dst)
path = dst
if (file_extension.lower() != ".wav"):
os.remove(temp_audio.name)
print("File name : " + str(path))
# strCovert = "ffmpeg -i "+"/transcribe_tmp/tmpbh97i2v0.webm" +" -c:a pcm_f32le "+/transcribe_tmp/ou2t.wav"
choose = 1
try:
choose = int(request.form['model'])
except:
pass
global model, model2, model3
runingModel = model
if (choose == 2):
runingModel = model2
print("Using model 2")
if (choose == 3):
runingModel = model3
print("Using model 3")
transcription, _ = run_transcribe(audio_path=path,
spect_parser=spect_parser,
model=runingModel,
decoder=decoder,
device=device,
use_half=True)
res['status'] = 200
res_text = ""
if (len(transcription) > 0):
res_text = transcription[0][0]
res['total'] = len(transcription[0])
else:
res_text = transcription
res['total'] = len(transcription)
res['data'] = transcribe_comma.runTranscribe(
commo_model, dict_data, word_dict, char_dict, res_text)
res['path'] = path
transTxt = path.replace("wav", "txt")
with open(transTxt, "w") as textFile:
textFile.write(res['data'])
logging.info('Success transcript')
logging.debug(res)
#os.remove(dst2)
except Exception as exx:
res['status'] = 403
res['data'] = str(exx)
t1 = time.time()
total = t1 - t0
targetString = ""
wer = 100
cer = 0
try:
targetString = request.form['targetString']
wer = werPecentage(targetString, res_text)
cer = cerPecentage(targetString, res_text)
except:
wer = 0
er = 0
res['seconds'] = total
res['wer'] = round(wer, 3)
res['cer'] = round(cer, 3)
return res
def transcribe_file():
if request.method == 'POST':
res = {}
res['total'] = 0
res['seconds'] = 0
t0 = time.time()
transTxt = ""
if 'file' not in request.files and 'url' not in request.form:
res['code'] = 403
res['data'] = "Missed audio files or url of mp3 file."
return jsonify(res)
try:
#đây là trường hợp có 1 trong 2 tham số 'file' và 'url', hoặc có cả 2
#***TH1 có tham số 'file'
file_extension = ""
path = "" #đường dẫn lưu file âm thanh ở server cần nhận dạng
if ('file' in request.files):
file = request.files['file']
filename = file.filename
_, file_extension = os.path.splitext(filename)
if file_extension.lower() not in ALLOWED_EXTENSIONS:
res['code'] = 403
res['data'] = "{} is not supported format.".format(
file_extension)
return jsonify(res)
with NamedTemporaryFile(prefix="product_",
suffix=file_extension,
dir='/work/dataset_product/wav',
delete=False) as temp_audio:
file.save(
temp_audio.name
) #lưu file cần nhận dạng vào đường dẫn temp_audio.name
path = temp_audio.name
elif ('url'
in request.form): #***TH2 ko có 'file', có tham số 'url'
url = request.form['url'] #đường dẫn mp3 hoặc video
#tải về
folder = 'work/dataset_recording/wav'
absolute_path = download_file(url, folder)
#nếu là mp3
if (absolute_path).endswith('.mp3'):
path = absolute_path
#nếu là mp4 : tách audio trong video
elif (absolute_path).endswith('.mp4'):
my_clip = mp.VideoFileClip(absolute_path)
filename = os.path.splitext(absolute_path)[0]
path = os.path.join(folder, filename + '.mp3')
my_clip.audio.write_audiofile(path)
#nếu ko phải
else:
res['code'] = 403
res['data'] = "Extension is not supported."
return jsonify(res)
#Chuyển đổi file âm thanh đúng định dạng wav, trả về đường dẫn wav sau chuyển đổi, file wav mới có tên giống file cũ
path = ConvertAudioToWav(path)
print("File name : " + str(path))
# strCovert = "ffmpeg -i "+"/transcribe_tmp/tmpbh97i2v0.webm" +" -c:a pcm_f32le "+/transcribe_tmp/ou2t.wav"
choose = 1
try:
choose = int(request.form['model'])
except:
pass
global model, model2, model3
runingModel = model
if (choose == 2):
runingModel = model2
print("Using model 2")
if (choose == 3):
runingModel = model3
print("Using model 3")
transcription, transcriptionGreedy, _, _ = run_transcribe(
audio_path=path,
spect_parser=spect_parser,
model=runingModel,
decoder=decoder,
device=device,
use_half=True)
res['status'] = 200
res_text = ""
if (len(transcription) > 0):
res_text = transcription[0][0]
res['total'] = len(transcription[0])
else:
res_text = transcription
res['total'] = len(transcription)
res['data'] = transcribe_comma.runTranscribe(
commo_model, dict_data, word_dict, char_dict, res_text)
res['path'] = path
res['greedy'] = transcribe_comma.runTranscribe(
commo_model, dict_data, word_dict, char_dict,
transcriptionGreedy[0][0])
transTxt = path.replace("wav", "txt")
with open(transTxt, "w") as textFile:
textFile.write(res['data'])
logging.info('Success transcript')
logging.debug(res)
except Exception as exx:
res['status'] = 403
res['data'] = "Không thể nhận dạng\n" + str(exx)
t1 = time.time()
total = t1 - t0
targetString = ""
wer = 100
cer = 0
try:
targetString = request.form['targetString']
wer = werPecentage(targetString, res_text)
cer = cerPecentage(targetString, res_text)
except:
wer = 0
er = 0
res['seconds'] = total
res['wer'] = round(wer, 3)
res['cer'] = round(cer, 3)
return res