로컬에서 Ollama설치와 STT, TTS 실행해 보기

 

인식도 잘되고 하구만

 

 

import requests
import json
from gtts import gTTS
import playsound as ps
import speech_recognition as sr

"""
1. 로컬에 ollama와 llama3.1:8b 설치
2. STT를 ko로 고정
3. TTS를 ko로 고정


[의존성 관련]

pip install gtts
pip install ollama
pip install playsound
pip3 install PyObjC
pip install SpeechRecognition

pip install -U pip
pip install -U pyaudio
"""

def speak_text(text):
    tts = gTTS(text, lang='ko', slow=False)
    tts.save("response.mp3")
    # os.system("mpg321 response.mp3")
    ps.playsound("response.mp3")


def stt():
    r = sr.Recognizer()
    with sr.Microphone() as source:
        print("Say something! : ")
        audio = r.listen(source)

    return r.recognize_google(audio, language='ko-KR')


def ollama_chat(text):
    url = "http://localhost:11434/api/generate"
    data = {
        "model": "llama3.1:8b",
        "prompt": text
    }

    header = {'Content-Type': 'application/json'}

    response = requests.post(url, json=data, headers=header)

    if response.status_code == 200:
        print( response.content.decode().strip().split('\n') )
        json_obj = response.content.decode().strip().split('\n')

        res_data =[json.loads(obj) for obj in json_obj]
        res_text = ''

        for item in res_data:
            res_text += item['response']

        # print(res_text)
        return (True, res_text)

    else:
        # print('Error:', response.status_code, response.text)
        return (False, response.text)

    return (False, '')


if __name__ == "__main__":
    # 음성 입력
    message = stt()
    print('질문: ' , message)

    # 생각 답변
    (status, text) = ollama_chat(message)
    print('답변: ', text)

    # 음성 출력
    speak_text(text)

 

 

 

1. 로컬에 Ollama설치 후 server를 띄우면 이렇게 확인 가능함.

 

2. Python에서 의존성 설치 (venv환경에서 진행)

- 필요에 따른 설치 

pip install gtts
pip install ollama
pip install playsound
pip3 install PyObjC
pip install SpeechRecognition

pip install -U pip
pip install -U pyaudio