Update app.py

app.py

@@ -8,6 +8,7 @@ model = torch.hub.load("PeterL1n/RobustVideoMatting", "mobilenetv3") # or "resne
 convert_video = torch.hub.load("PeterL1n/RobustVideoMatting", "converter")
 
 def inference(video):
+  os.system('rm output.mp4')
   os.system('ffmpeg -ss 00:00:00 -i '+ video +' -to 00:00:05 -c copy output.mp4')
   convert_video(
       model,                           # The loaded model, can be on any device (cpu or cuda).