Update app.py

app.py

@@ -7,7 +7,7 @@ client = InferenceClient("google/gemma-2-27b-it")
 
 def generate_text(messages):
     generated = ""
-    for token in client.chat_completion(messages, max_tokens=100,stream=True):
+    for token in client.chat_completion(messages, max_tokens=50,stream=True):
         content = (token.choices[0].delta.content)
         generated+=content
         print(generated)
@@ -40,11 +40,17 @@ head = '''
 '''
 with gr.Blocks(title="LLM with TTS",head=head) as demo:
     gr.Markdown("## Please be patient, the first response may have a delay of up to 20 seconds while loading.")
-    gr.Markdown("**Qwen2.5-0.5B-Instruct/LJSpeech-q8**.LLM and TTS models will change without notice.")
+    gr.Markdown("**gemma-2-27b-it/LJSpeech**.LLM and TTS models will change without notice.")
+    
     js = """
+    function replaceSpecialChars(text) {
+    const pattern = /[^a-zA-Z0-9,.!?-_']/g;
+    return text.replace(pattern, ' ');
+    }
+
     function(chatbot){
     text = (chatbot[chatbot.length -1])["content"]
-    window.MatchaTTSEn(text)
+    window.MatchaTTSEn(replaceSpecialChars(text))
     }
     """
     chatbot = gr.Chatbot(type="messages")