app.py

# Import libraries
import os
import gradio as gr
import torch
import soundfile as sf
import numpy as np
from PIL import Image
import torch.nn.functional as F
import logging
from scipy.io.wavfile import write as write_wav
from scipy import signal
from moviepy.editor import VideoFileClip, AudioFileClip
from transformers import AutoProcessor, AutoModelForAudioGeneration
import requests  # Add this line

# Set up logging for better debug tracking
logging.basicConfig(level=logging.DEBUG, format="%(asctime)s - %(levelname)s - %(message)s")
logger = logging.getLogger()

# Download Places365 class labels
try:
    logging.info("Downloading Places365 class labels...")
    url = "http://places2.csail.mit.edu/models_places365/categories_places365.txt"
    response = requests.get(url)
    with open("categories_places365.txt", "wb") as f:
        f.write(response.content)
    logging.info("Places365 class labels downloaded successfully.")
except Exception as e:
    logging.error(f"Error downloading Places365 class labels: {e}")
    raise

# Load Places365 model for scene detection (on CPU to save GPU memory)
try:
    logging.info("Loading Places365 model for scene detection...")
    places365 = torch.hub.load('pytorch/vision:v0.10.0', 'resnet18', pretrained=True)
    places365.eval()
    places365.to("cpu")  # Move to CPU
    logging.info("Places365 model loaded successfully.")
except Exception as e:
    logging.error(f"Error loading Places365 model: {e}")
    raise

# Load Places365 class labels
with open("categories_places365.txt", "r") as f:
    SCENE_CLASSES = [line.strip().split(" ")[0][3:] for line in f.readlines()]

# Load AudioGen Medium and MusicGen Medium models
try:
    logging.info("Loading AudioGen Medium and MusicGen Medium models...")
    audiogen_processor = AutoProcessor.from_pretrained("facebook/audiogen-medium")
    audiogen_model = AutoModelForAudioGeneration.from_pretrained("facebook/audiogen-medium")
    musicgen_processor = AutoProcessor.from_pretrained("facebook/musicgen-medium")
    musicgen_model = AutoModelForAudioGeneration.from_pretrained("facebook/musicgen-medium")
    
    # Move models to GPU if available
    device = "cuda" if torch.cuda.is_available() else "cpu"
    audiogen_model.to(device)
    musicgen_model.to(device)
    logging.info("AudioGen Medium and MusicGen Medium models loaded successfully.")
except Exception as e:
    logging.error(f"Error loading AudioGen/MusicGen models: {e}")
    raise

# Rest of the code remains the same...