app.py

import gradio as gr

import os
import wave
import json
import requests
import urllib.request
import speech_recognition
from vosk_tts import Model, Synth
from vosk import Model, KaldiRecognizer
from scipy.io.wavfile import write
from pydub import AudioSegment
from pydub.playback import play


PATH_TO_MODEL = "vosk-model-tts-ru-0.4-multi"
PATH_TO_OUTPUT = "C:/Users/user/Desktop/deepfake_sirius/materials/audio" #TODO: IT


k = "sk-YOVNQzHmpga9My3dwlSo9BQN907TuPZQXcHn50ztigTwm3I2"
files = [
    ("input_face", open("C:\\Users\\user\\Desktop\\deepfake_sirius\\materials\\scale_1200.jpg", "rb")), #TODO: IT
    ("input_audio", open("C:\\Users\\user\\Desktop\\deepfake_sirius\\materials\\audio\\output.wav", "rb")), #TODO: IT
]
payload = {}


class VoiceGenerator:
    def __init__(self):
        self.model = Model(model_path=PATH_TO_MODEL)
    def generate(self, text, file_name='output.wav'):
        synth = Synth(self.model)
        path = os.path.join(PATH_TO_OUTPUT, file_name)
        synth.synth(text, path)
        return path


def record_and_recognize_audio(file_path):
    with speech_recognition.AudioFile(file_path) as source:
        audio = recognizer.record(source)
    try:
        recognized_data = recognizer.recognize_google(audio, language="ru").lower()
    except speech_recognition.UnknownValueError:
        pass
    except speech_recognition.RequestError:
        pass
    return recognized_data


def ask(request):
    instruction = """
    Ответь на запрос так, как ответил бы на него Павел Воля. Используй данные из биографии Павла Воли, если это потребуется. Отвечай на запрос в его стиле. Ответ должен содержать не болеее 10 предложений.
    """
    result = requests.post(
        url='https://llm.api.cloud.yandex.net/llm/v1alpha/instruct',
        headers={
            "Authorization": "Api-Key AQVNyVqBi-XoJ1cAo7VIxq6ztgXm3owqowtso5Qb",
        },
        json={
            "model": "general",
            "instruction_text": instruction,
            "request_text": request,
            "generation_options": {
                "max_tokens": 1500,
                "temperature": 0.5
            }
        }
    )
    data = json.loads(result.text)
    return(data['result']['alternatives'][0]['text'])


def result(audio):
    sr, data = audio
    print('sr:', sr, 'data:', data)
    #return os.path.join(os.path.abspath(''), "video_sample.mp4")
    write('voice_input.wav', sr, data)
    return os.path.join(os.path.abspath(''), "voice_input.wav")

demo = gr.Interface(
    result,
    gr.Audio(sources=["microphone"]),
    "audio", #playable_video
)

demo.launch()