update detail description.

app.py

@@ -150,7 +150,10 @@ iface = gr.Interface(
     article="- The motivation of the ConST model is to use the contrastive learning method to learn similar representations for semantically similar speech and text, " \
             "thus leveraging MT to help improve ST performance. \n"
             "- The models you are experiencing are trained based on the MuST-C dataset (https://ict.fbk.eu/must-c/), " \
-            "which only contains about 250k parallel data at each translation direction. \n"
+            "which only contains about 250k parallel data at each translation direction. "
+            "The translation performance of these language directions varies from 20-30+ BLEU, "
+            "so it is normal to find some flaws in the translation, and we are trying to improve the models, "
+            "such as training on larger datasets and developing more advanced algorithms.\n"
             "- If you want to know how to train the models, you may refer to https://github.com/ReneeYe/ConST.",
     theme="peach",
 )