src/infer.py

from typing import *
from src import utils
from transformers import WhisperProcessor, WhisperForConditionalGeneration
import whisper

def predict(model_name, language, mic_audio=None, audio=None):
    if mic_audio is not None:
        voice = mic_audio
    elif audio is not None:
        voice = audio
    else:
        return "(please provide audio)"
    
    voice = utils.preprocess_audio(voice)
    
    model = whisper.load_model(model_name)
    result = model.transcribe(voice, language=language)
    return result["text"]