Update app.py

app.py

@@ -118,7 +118,7 @@ def predict_and_save(process_video):
     
     # Collect timestamps and texts
     transcript = []
-    for i, (start, end) in enumerate(sample['net_input']['audio_lengths'], 1):
+    for i, (start, end) in enumerate(sample['net_input']['video_lengths'], 1):
         start_time = float(start) / 16_000
         end_time = float(end) / 16_000
         text = hypo[i].strip()
@@ -130,6 +130,7 @@ def predict_and_save(process_video):
     
     return hypo
 
+
 def preprocess_video(input_video_path):
     if torch.cuda.is_available():
         detector = dlib.cnn_face_detection_model_v1(face_detector_path)