audio2text/a2t.py

import numpy as np

from .init import pipe

TASK = "transcribe"
BATCH_SIZE = 16

class A2T:
    def __init__(self, mic):
        self.mic = mic

    def __transcribe(self, inputs, task: str = None):
        if inputs is  None:
            print("Inputs None")

        transcribed_text = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs={"task": task}, return_timestamps=True)["text"]
        print("transcribed_text : ", transcribed_text)
        return transcribed_text
        
    def predict(self):
        try:
            if self.mic is not None:
                chunk = self.mic.get_array_of_samples()
                audio = np.array(chunk)
                print("audio : ", audio)
            else:
                return "please provide audio"

            if isinstance(audio , np.ndarray):
                return self.__transcribe(inputs=audio, task=TASK)
            else:
                return "Audio is not np array"
                
        except Exception as e:
            print("Predict error", e)
            return "Oops some kinda error"