import numpy as np from .init import pipe TASK = "transcribe" BATCH_SIZE = 8 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, "language": "english"}) print("transcribed_text : ", transcribed_text) return transcribed_text["text"] def __preprocces(self, raw: np.ndarray): chunk = raw.astype(np.float32) / 32768.0 return chunk def predict(self): try: if self.mic is not None: chunk = self.mic.get_array_of_samples() chunk = np.array(chunk, dtype=np.int16) audio = self.__preprocces(chunk) print(f"audio : {audio} \n shape : {audio.shape} \n max : {np.max(audio)}") else: raise Exception("please provide audio") if isinstance(audio , np.ndarray): inputs = {"sampling_rate": 16000, "raw": audio} return self.__transcribe(inputs=inputs, task=TASK) else: raise Exception("Audio is not np array") except Exception as e: return f"Oops some kinda error : {e}"