diff --git a/src/server/bot/core/component_factory.py b/src/server/bot/core/component_factory.py
index c7be831..921c92b 100644
--- a/src/server/bot/core/component_factory.py
+++ b/src/server/bot/core/component_factory.py
@@ -23,6 +23,7 @@
 from ..components.llm_tools.end_conversation_handler import EndConversationHandler
 from ..processors.speech.lipsync_processor import LipsyncProcessor
 from ..transport.custom_services.kokoro_service import KokoroTTSService
+from ..transport.custom_services.emoji_text_filter import EmojiTextFilter
 from ..utils.device_utils import get_best_device
 from ..transport.custom_services.ollama_service import CustomOLLamaLLMService
 from ..components.memory import MemoryHandler
@@ -178,6 +179,10 @@ def build_instruction(self) -> str:
             "discourse markers (you know, I mean), and brief pauses (uh, um). "
             "Use sparingly; never in numbers or names.\n"
         )
+        instruction += (
+            "Important: Do not use emojis or any special unicode symbols. "
+            "Your responses are converted directly to audio — emojis will be read aloud as their names.\n"
+        )
 
         return instruction.strip()
 
@@ -377,20 +382,23 @@ def _build_tts_service(self) -> Optional[object]:
         if not self.tts_type:
             return None
 
+        emoji_filter = EmojiTextFilter()
         if self.tts_type == "openai":
             return OpenAITTSService(
                 voice=self._get_voice_for_openai(),
                 model=(self.tts_params or {}).get("model", "gpt-4o-mini-tts"),
+                text_filters=[emoji_filter],
             )
         elif self.tts_type == "elevenlabs":
             return ElevenLabsTTSService(
                 api_key=os.getenv("ELEVENLABS_API_KEY"),
                 voice_id=self._get_voice_id_for_elevenlabs(),
                 model="eleven_flash_v2_5",
+                text_filters=[emoji_filter],
             )
         elif self.tts_type == "kokoro":
             device = get_best_device()
             return KokoroTTSService(
-                voice=self._get_voice_id_for_kokoro(), device=device
+                voice=self._get_voice_id_for_kokoro(), device=device, text_filters=[emoji_filter]
             )
         return None
diff --git a/src/server/bot/transport/custom_services/emoji_text_filter.py b/src/server/bot/transport/custom_services/emoji_text_filter.py
new file mode 100644
index 0000000..fe2e091
--- /dev/null
+++ b/src/server/bot/transport/custom_services/emoji_text_filter.py
@@ -0,0 +1,39 @@
+import re
+from typing import Any, Mapping
+
+from pipecat.utils.text.base_text_filter import BaseTextFilter
+
+_EMOJI_RE = re.compile(
+    "["
+    "\U0001F600-\U0001F64F"
+    "\U0001F300-\U0001F5FF"
+    "\U0001F680-\U0001F6FF"
+    "\U0001F1E0-\U0001F1FF"
+    "\U00002600-\U000026FF"
+    "\U00002700-\U000027BF"
+    "\U0000FE00-\U0000FE0F"
+    "\U0001F900-\U0001F9FF"
+    "\U0001FA70-\U0001FAFF"
+    "\U00002300-\U000023FF"
+    "\U00002B50-\U00002B55"
+    "\U0001F004"
+    "\U0001F0CF"
+    "]+",
+    flags=re.UNICODE,
+)
+
+
+class EmojiTextFilter(BaseTextFilter):
+    """Strips emoji characters from text before TTS synthesis."""
+
+    async def update_settings(self, settings: Mapping[str, Any]):
+        pass
+
+    async def filter(self, text: str) -> str:
+        return _EMOJI_RE.sub("", text).strip()
+
+    async def handle_interruption(self):
+        pass
+
+    async def reset_interruption(self):
+        pass