Update app.py

app.py

@@ -284,7 +284,7 @@ def create_interface():
             if input_mode == "Text":
                 return input_text, None, None
             elif input_mode == "Image":
-                return None, image_input, None
+                return input_text, image_input, None
             else:  # Voice
                 return None, None, audio_input