File size: 1,687 Bytes
003e418 d5436e0 4d6b36a 003e418 d5436e0 ecf0dad 003e418 d5436e0 ecf0dad 003e418 d5436e0 003e418 d5436e0 003e418 d5436e0 003e418 ecf0dad 003e418 ecf0dad |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 |
import numpy as np
import librosa
import io
from typing import Optional
from .config import pipe
TASK = "transcribe"
BATCH_SIZE = 8
class A2T:
def __init__(self, mic):
self.mic = mic
def __generate_text(self, inputs, task: Optional[str] = None) -> str:
if inputs is None:
raise Exception("Inputs is None")
transcribed_text = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs={"task": task}, return_timestamps=True)["text"]
return transcribed_text
@staticmethod
def __preprocess(raw: bytes) -> np.ndarray:
print(f"Raw type: {type(raw)}")
if not isinstance(raw, bytes):
raise ValueError("Expected raw audio data as bytes")
try:
chunk = io.BytesIO(raw)
print(f"Chunk type: {type(chunk)}")
audio, sample_rate = librosa.load(chunk, sr=16000)
print(f"Sample rate : {sample_rate}")
return audio
except Exception as e:
print(f"Error loading audio: {e}")
def predict(self) -> str:
try:
if self.mic is not None:
raw = self.mic
audio = self.__preprocess(raw=raw)
print(f"audio type : {type(audio)} \n shape : {audio.shape} \n audio max value : {np.max(audio)}")
else:
raise Exception("please provide audio")
if isinstance(audio, np.ndarray):
return self.__generate_text(inputs=audio, task=TASK)
else:
raise ValueError("Audio is not np array")
except Exception as e:
print(f"Oops some kinda error : {e}") |