diff --git a/main.py b/main.py
index 0c79409ee..93602eb65 100644
--- a/main.py
+++ b/main.py
@@ -74,10 +74,10 @@ def handle_text_message(event):
 
         elif text.startswith('/圖像'):
             prompt = text[3:].strip()
+            memory.append(user_id, 'user', prompt)
             is_successful, response, error_message = model_management[user_id].image_generations(prompt)
             if not is_successful:
                 raise Exception(error_message)
-            memory.append(user_id, 'user', prompt)
             msg = ImageSendMessage(
                 original_content_url=response,
                 preview_image_url=response
@@ -85,17 +85,18 @@ def handle_text_message(event):
             memory.append(user_id, 'assistant', response)
 
         else:
+            memory.append(user_id, 'user', text)
             is_successful, response, error_message = model_management[user_id].chat_completions(memory.get(user_id), os.getenv('OPENAI_MODEL_ENGINE'))
             if not is_successful:
                 raise Exception(error_message)
-            memory.append(user_id, 'user', text)
             role, response = get_role_and_content(response)
             msg = TextSendMessage(text=response)
             memory.append(user_id, role, response)
 
     except ValueError:
-        msg = TextSendMessage(text='Token 無效，請重新註冊，注意格式有空格，格式為 /註冊 sk-xxxxx')
+        msg = TextSendMessage(text='Token 無效，請重新註冊，格式為 /註冊 sk-xxxxx')
     except Exception as e:
+        memory.remove(user_id)
         msg = TextSendMessage(text=str(e))
     line_bot_api.reply_message(event.reply_token, msg)
 
@@ -114,18 +115,19 @@ def handle_audio_message(event):
             raise ValueError('Invalid API token')
         else:
             transciption, error_message = model_management[user_id].audio_transcriptions(input_audio_path, 'whisper-1')
+            memory.append(user_id, 'user', transciption)
             if error_message:
                 raise Exception(error_message)
             is_successful, response, error_message = model_management[user_id].chat_completions(memory.get(user_id), 'gpt-3.5-turbo')
             if not is_successful:
                 raise Exception(error_message)
-            memory.append(user_id, 'user', transciption)
             role, response = get_role_and_content(response)
             memory.append(user_id, role, response)
             msg = TextSendMessage(text=response)
     except ValueError:
         msg = TextSendMessage(text='請先註冊你的 API Token，格式為 /註冊 [API TOKEN]')
     except Exception as e:
+        memory.remove(user_id)
         msg = TextSendMessage(text=str(e))
     os.remove(input_audio_path)
     line_bot_api.reply_message(event.reply_token, msg)
diff --git a/src/models.py b/src/models.py
index 973d78d62..87acc3d67 100644
--- a/src/models.py
+++ b/src/models.py
@@ -6,10 +6,10 @@ class ModelInterface:
     def check_token_valid(self) -> bool:
         pass
 
-    def chat_completions(self, messages: List[Dict]) -> str:
+    def chat_completions(self, messages: List[Dict], model_engine: str) -> str:
         pass
 
-    def audio_transcriptions(self, file) -> str:
+    def audio_transcriptions(self, file, model_engine: str) -> str:
         pass
 
     def image_generations(self, prompt: str) -> str:
@@ -62,4 +62,4 @@ def image_generations(self, prompt: str) -> str:
             "n": 1,
             "size": "512x512"
         }
-        return self._request('/images/generations', body=json_body)
+        return self._request('POST', '/images/generations', body=json_body)
diff --git a/src/utils.py b/src/utils.py
index dccad92d9..032ffdf25 100644
--- a/src/utils.py
+++ b/src/utils.py
@@ -4,8 +4,8 @@
 t2s_converter = opencc.OpenCC('t2s.json')
 
 
-def get_role_and_content(response):
+def get_role_and_content(response: str):
     role = response['choices'][0]['message']['role']
     content = response['choices'][0]['message']['content'].strip()
-    response = s2t_converter.convert(content)
-    return role, response
+    content = s2t_converter.convert(content)
+    return role, content