mirror of
https://github.com/kimjammer/Neuro
synced 2025-07-08 19:04:24 -04:00
51 lines
1.5 KiB
Python
51 lines
1.5 KiB
Python
|
import requests
|
||
|
|
from RealtimeTTS import TextToAudioStream, CoquiEngine
|
||
|
|
from RealtimeSTT import AudioToTextRecorder
|
||
|
|
|
||
|
|
def process_text(text):
|
||
|
|
print(text)
|
||
|
|
user_message = text
|
||
|
|
history.append({"role": "user", "content": user_message})
|
||
|
|
data = {
|
||
|
|
"mode": "chat-instruct",
|
||
|
|
"character": "AI",
|
||
|
|
"messages": history
|
||
|
|
}
|
||
|
|
|
||
|
|
response = requests.post(url, headers=headers, json=data, verify=False)
|
||
|
|
assistant_message = response.json()['choices'][0]['message']['content']
|
||
|
|
history.append({"role": "assistant", "content": assistant_message})
|
||
|
|
print(assistant_message)
|
||
|
|
stream.feed(assistant_message)
|
||
|
|
stream.play()
|
||
|
|
|
||
|
|
if __name__ == '__main__':
|
||
|
|
engine = CoquiEngine(
|
||
|
|
use_deepspeed=True,
|
||
|
|
voice="./voices/David_Attenborough CC3.wav"
|
||
|
|
)
|
||
|
|
stream = TextToAudioStream(engine)
|
||
|
|
|
||
|
|
url = "http://127.0.0.1:5000/v1/chat/completions"
|
||
|
|
headers = {"Content-Type": "application/json"}
|
||
|
|
history = []
|
||
|
|
|
||
|
|
recorder_config = {
|
||
|
|
'spinner': False,
|
||
|
|
'model': 'tiny.en',
|
||
|
|
'language': 'en',
|
||
|
|
'silero_sensitivity': 0.4,
|
||
|
|
'silero_use_onnx': True,
|
||
|
|
'webrtc_sensitivity': 2,
|
||
|
|
'post_speech_silence_duration': 0.4,
|
||
|
|
'min_length_of_recording': 0,
|
||
|
|
'min_gap_between_recordings': 0,
|
||
|
|
'enable_realtime_transcription': True,
|
||
|
|
'realtime_processing_pause': 0.2,
|
||
|
|
'realtime_model_type': 'tiny.en'
|
||
|
|
}
|
||
|
|
|
||
|
|
|
||
|
|
with AudioToTextRecorder(**recorder_config) as recorder:
|
||
|
|
while True:
|
||
|
|
recorder.text(process_text)
|