diff --git a/backend/main/xiaozhi-server/core/providers/llm/antaf/antaf.py b/backend/main/xiaozhi-server/core/providers/llm/antaf/antaf.py
index b011e3a..438e95d 100644
--- a/backend/main/xiaozhi-server/core/providers/llm/antaf/antaf.py
+++ b/backend/main/xiaozhi-server/core/providers/llm/antaf/antaf.py
@@ -89,6 +89,8 @@ class LLMProvider(LLMProviderBase):
             yield "抱歉，我没有收到您的问题。"
             return
 
+        # 追加简短回答提示，避免阿福回复过长导致TTS排队卡顿
+        query = query + "（请用2-3句话简短回答）"
         logger.bind(tag=TAG).info(f"AntafLLM 请求: {query[:50]}...")
 
         try:
diff --git a/backend/main/xiaozhi-server/core/providers/tts/sherpa_tts.py b/backend/main/xiaozhi-server/core/providers/tts/sherpa_tts.py
index a997ac6..de0efc8 100644
--- a/backend/main/xiaozhi-server/core/providers/tts/sherpa_tts.py
+++ b/backend/main/xiaozhi-server/core/providers/tts/sherpa_tts.py
@@ -70,8 +70,7 @@ class TTSProvider(TTSProviderBase):
         return wav_io.getvalue()
 
     async def text_to_speak(self, text, output_file):
-        loop = asyncio.get_event_loop()
-        wav_data = await loop.run_in_executor(None, self._generate_wav, text)
+        wav_data = self._generate_wav(text)
 
         if output_file:
             with open(output_file, "wb") as f: