#!/usr/bin/env python3

import json
import os
import pyaudio
from vosk import Model, KaldiRecognizer


lang = en-US
DEBUG = True


class Vosk:
    def __init__(self, language):
        if not os.path.exists("models/vosk/" + language):
            print("Please download the model from https://alphacephei.com/vosk/models and unpack as "
                  "'" + language + " ' in 'models/vosk'.")
            exit(1)

        model = Model("models/vosk/" + language)
        self.rec = KaldiRecognizer(model, 16000)

        p = pyaudio.PyAudio()
        self.stream = p.open(format=pyaudio.paInt16,
                             channels=1,
                             rate=16000,
                             input=True,
                             frames_per_buffer=8000)
        self.stream.start_stream()

    def run(self):
        print("Listening...")
        while True:
            data = self.stream.read(4000)
            if len(data) == 0:
                break
            if self.rec.AcceptWaveform(data):
                res = json.loads(self.rec.FinalResult())
                if DEBUG:
                    print("Text:", res['text'])

                res = json.loads(self.rec.Result())
                if DEBUG:
                    print(res['text'])

        res = json.loads(self.rec.FinalResult())
        if DEBUG:
            print("Listened: " + res['text'])
        return res['text']


if __name__ == '__main__':
    stt = Vosk(lang)
    print(stt.run())