Update module

app.py

@@ -60,6 +60,7 @@ import os
 import tempfile
 from ultralytics import YOLO
 from moviepy import ImageSequenceClip
+from scipy.io import wavfile
 
 # Load face detector
 model = YOLO("yolov8n-face.pt").cuda()  # assumes CUDA available
@@ -184,7 +185,7 @@ def extract_faces(video_file):
 
     # ------------------------------- #
     # AVSE models
-    noisy = self.load_wav(audio_path)
+    noisy = wavfile.read(audio_path)[1].astype(np.float32) / (2 ** 15)
 
     vr = VideoReader(output_path, ctx=cpu(0))
     frames = vr.get_batch(list(range(len(vr)))).asnumpy()