def build_first_request(sample_rate_hertz, num_channels):
request = stt_pb2.StreamingRecognizeRequest()
request.streaming_config.config.encoding = stt_pb2.AudioEncoding.LINEAR16
request.streaming_config.config.sample_rate_hertz = sample_rate_hertz
request.streaming_config.config.num_channels = num_channels
request.streaming_config.config.enable_automatic_punctuation = True
return request
def generate_requests(args, agi):
try:
yield build_first_streaming_recognition_request(args)
i = 0
while True:
i = i + 1
if agi != None:
data = os.read(AUDIO_FD, 8000)
else:
data = bytes(160)
if i == 1: # drop initial data
continue
#if agi != None:
# agi.verbose("Frame: %d(%d): %s" % (i, len(data), binascii.hexlify(data)))
#else:
# print("Frame: %d(%d): %s" % (i, len(data), binascii.hexlify(data)))
if not data:
break
request = stt_pb2.StreamingRecognizeRequest()
request.audio_content = data
if agi == None and i > 20:
sys.exit()
yield request
except:
logger.exception("Got exception in generate_requests")
raise
def build_first_request(sample_rate_hertz, num_channels):
request = stt_pb2.StreamingRecognizeRequest()
request.streaming_config.config.encoding = stt_pb2.AudioEncoding.LINEAR16
request.streaming_config.config.sample_rate_hertz = sample_rate_hertz
request.streaming_config.config.num_channels = num_channels
request.streaming_config.config.profanity_filter = True
return request
def generate_requests(context):
try:
with wave.open("../../audio/stierlitz.wav") as f:
yield build_first_request(f.getframerate(), f.getnchannels(),
context)
frame_samples = f.getframerate() // 10 # Send 100ms at a time
for data in iter(lambda: f.readframes(frame_samples), b''):
request = stt_pb2.StreamingRecognizeRequest()
request.audio_content = data
yield request
# Sending 1 second of silence
for i in range(10):
request = stt_pb2.StreamingRecognizeRequest()
request.audio_content = bytes(frame_samples)
yield request
except Exception as e:
print("Got exception in generate_requests", e)
raise
def build_first_request(sample_rate_hertz, num_channels):
request = stt_pb2.StreamingRecognizeRequest()
request.streaming_config.config.encoding = stt_pb2.AudioEncoding.LINEAR16
request.streaming_config.config.sample_rate_hertz = sample_rate_hertz
request.streaming_config.config.num_channels = num_channels
request.streaming_config.config.vad_config.min_speech_duration = 1.0
request.streaming_config.config.vad_config.max_speech_duration = 30.0
request.streaming_config.config.vad_config.silence_duration_threshold = 3.0
request.streaming_config.config.vad_config.silence_prob_threshold = 0.2
return request
def build_first_request(sample_rate_hertz, num_channels, context):
request = stt_pb2.StreamingRecognizeRequest()
request.streaming_config.config.encoding = stt_pb2.AudioEncoding.LINEAR16
request.streaming_config.config.sample_rate_hertz = sample_rate_hertz
request.streaming_config.config.num_channels = num_channels
# ВАЖНО!
# Не рекомендуется:
# - добавление слов короче 5 символов
# - выставление score в слишком большое значение
request.streaming_config.config.speech_contexts.append(context)
return request
def generate_requests():
try:
yield build_first_request()
with open("../../audio/sample_3.al", "rb") as f:
for data in iter(lambda: f.read(800), b''): # Send 100ms at a time
request = stt_pb2.StreamingRecognizeRequest()
request.audio_content = data
yield request
except Exception as e:
print("Got exception in generate_requests", e)
raise
def generate_requests(path):
try:
with wave.open(path) as f:
yield build_first_request(f.getframerate(), f.getnchannels())
frame_samples = f.getframerate() // 10 # Send 100ms at a time
for data in iter(lambda: f.readframes(frame_samples), b''):
request = stt_pb2.StreamingRecognizeRequest()
request.audio_content = data
yield request
except Exception as e:
print("Got exception in generate_requests", e)
raise
def stt_generate_requests(args, reader):
try:
yield build_first_streaming_recognition_request(args)
while True:
data = reader.read()
if not data:
break
request = stt_pb2.StreamingRecognizeRequest()
request.audio_content = data
yield request
except:
logger.exception("Got exception in stt_generate_requests")
raise
def generate_requests():
try:
fname = "../../audio/sample_3.mp3"
info = MP3(fname).info
yield build_first_request(info.sample_rate, info.channels)
with open(fname, "rb") as f:
for data in iter(lambda: f.read(4096),
b''): # Send 4096 bytes at a time
request = stt_pb2.StreamingRecognizeRequest()
request.audio_content = data
yield request
except Exception as e:
print("Got exception in generate_requests", e)
raise
def requests(request):
try:
yield request
pyaudio_lib = pyaudio.PyAudio()
f = pyaudio_lib.open(input=True,
channels=1,
format=pyaudio.paInt16,
rate=16000)
for data in iter(lambda: f.read(800), b''):
request = stt_pb2.StreamingRecognizeRequest()
request.audio_content = data
yield request
except Exception as e:
print("Got exception in generate_requests", e)
raise
def generate_requests():
try:
sample_rate_hertz, num_channels = 16000, 1
pyaudio_lib = pyaudio.PyAudio()
f = pyaudio_lib.open(input=True,
channels=num_channels,
format=pyaudio.paInt16,
rate=sample_rate_hertz)
yield build_first_request(sample_rate_hertz, num_channels)
for data in iter(lambda: f.read(800), b''): # Send 50ms at a time
request = stt_pb2.StreamingRecognizeRequest()
request.audio_content = data
yield request
except Exception as e:
print("Got exception in generate_requests", e)
raise
def __init__(self, recognize):
self._recognize = recognize
r = stt_pb2.StreamingRecognizeRequest()
r.streaming_config.config.encoding = stt_pb2.AudioEncoding.LINEAR16
r.streaming_config.config.sample_rate_hertz = 16000
r.streaming_config.config.num_channels = 1
r.streaming_config.config.enable_denormalization = True
r.streaming_config.config.enable_automatic_punctuation = True
r.streaming_config.config.vad_config.silence_duration_threshold = 1.20
r.streaming_config.single_utterance = True
r.streaming_config.config.speech_contexts.append(
stt_pb2.SpeechContext(phrases=[
stt_pb2.SpeechContextPhrase(text=text, score=10.0)
for text in self._recognize.context_words
]))
metadata = authorization_metadata(API_KEY, SECRET_KEY,
"tinkoff.cloud.stt")
stub = stt_pb2_grpc.SpeechToTextStub(
grpc.secure_channel(ENDPOINT, grpc.ssl_channel_credentials()))
self._responses = stub.StreamingRecognize(self.requests(r),
metadata=metadata)
def generate_requests():
try:
with wave.open("../../audio/sample_3.wav") as f:
frame_rate_is_valid(f.getframerate())
yield build_first_request(f.getframerate(), f.getnchannels())
frame_samples = f.getframerate() // 1000 * 60 # 60ms
opus_encoder = opuslib.Encoder(f.getframerate(), f.getnchannels(),
opuslib.APPLICATION_AUDIO)
for data in iter(lambda: f.readframes(frame_samples),
b''): # Send 60ms at a time
if len(
data
) < frame_samples * 2: # Padding last frame to closest allowed frame size
data = data.ljust(
get_padded_frame_size(frame_samples, f.getframerate())
* 2, b'\0')
request = stt_pb2.StreamingRecognizeRequest()
request.audio_content = opus_encoder.encode(
data,
len(data) >> 1)
yield request
except Exception as e:
print("Got exception in generate_requests", e)
raise
def build_first_request(sample_rate_hertz, num_channels):
request = stt_pb2.StreamingRecognizeRequest()
request.streaming_config.config.encoding = stt_pb2.AudioEncoding.MPEG_AUDIO
request.streaming_config.config.sample_rate_hertz = sample_rate_hertz
request.streaming_config.config.num_channels = num_channels
return request
def build_first_streaming_recognition_request(args):
request = stt_pb2.StreamingRecognizeRequest()
set_recognition_config(request.streaming_config.config, args)
request.streaming_config.interim_results_config.enable_interim_results = args.interim_results
request.streaming_config.single_utterance = args.single_utterance
return request
def build_first_request():
request = stt_pb2.StreamingRecognizeRequest()
request.streaming_config.config.encoding = stt_pb2.AudioEncoding.LINEAR16
request.streaming_config.config.sample_rate_hertz = 16000 # Not stored at raw ".s16" file
request.streaming_config.config.num_channels = 1 # Not stored at raw ".s16" file
return request
