def run(self):
stream = self.deepspeech_model.createStream()
i = 1
for _bytes, src in self.voiced_audio_generator():
if _bytes is not None:
stream.feedAudioContent(np.frombuffer(_bytes, np.int16))
self.store_audio_data(_bytes, "voiced")
self.store_audio_data(_bytes, str(i))
else:
text = stream.finishStream()
stream = self.deepspeech_model.createStream()
if text:
logger.debug(f"Transcript #{i}: {text}")
publish_message(self, Text(text), src)
# else:
# logger.debug(f"Transcript #{i} is empty!")
i += 1
def run(self):
audio = pyaudio.PyAudio()
stream = audio.open(
format=pyaudio.paInt16,
channels=1,
rate=SpeechToTextListenerTask.sample_rate,
input=True,
frames_per_buffer=self.frames_per_buffer,
)
while self.running:
try:
input_data = stream.read(self.frames_per_buffer)
except e:
logger.error(e)
self.running = False
break
finally:
output_data = self._process_audio_data(input_data)
publish_message(self, AudioFrame(output_data))
stream.stop_stream()
stream.close()
audio.terminate()
def process_text(self, text: Text):
if not text.value:
return
response = self.chatbot.get_response(text.value)
if response.text:
publish_message(self, Text(response.text))
def on_message(self, data):
logger.debug(f"Received {len(data)} bytes")
message = Message.from_bytes(data)
if message is not None:
publish_message(
registry.get_task(BoardWebSocketHandlerListenerTask), message)
def speak(self, text: Text) -> None:
if not text.value:
logger.warning("Cannot synthesize empty text")
data = self.synthesize(text.value)
publish_message(self, AudioFrame(data), text.src)