solved

app.py

@@ -33,6 +33,8 @@ hifigan_loaded = torch.load("hifigan_cpu.pt", map_location=torch.device('cpu'))
 
 # Conversion function
 def convert_speech(filename, progress=gr.Progress()):
+    if not filename:
+        raise ValueError("Please provide an audio")    
     progress(0, desc="Starting conversion")
 
     progress(0.1, desc="Loading audio")