diff --git a/src/palabra_ai/enum.py b/src/palabra_ai/enum.py
index 483b28f..dd9163b 100644
--- a/src/palabra_ai/enum.py
+++ b/src/palabra_ai/enum.py
@@ -7,6 +7,7 @@ class MessageType(StrEnum):
     VALIDATED_TRANSCRIPTION = "validated_transcription"
     PARTIAL_TRANSLATED_TRANSCRIPTION = "partial_translated_transcription"
     PIPELINE_TIMINGS = "pipeline_timings"
+    TTS_TEXT = "tts_text"
     _QUEUE_STATUS = "queue_status"  # For "es" messages
     _EMPTY = "empty"  # For empty {} messages
     _UNKNOWN = "unknown"  # For unrecognized message formats
@@ -19,6 +20,7 @@ class MessageType(StrEnum):
         MessageType.TRANSLATED_TRANSCRIPTION,
         MessageType.VALIDATED_TRANSCRIPTION,
         MessageType.PARTIAL_TRANSLATED_TRANSCRIPTION,
+        MessageType.TTS_TEXT,
     )
 }
 ALLOWED_MESSAGE_TYPES = {
diff --git a/src/palabra_ai/message.py b/src/palabra_ai/message.py
index b2a2c49..5434f84 100644
--- a/src/palabra_ai/message.py
+++ b/src/palabra_ai/message.py
@@ -116,6 +116,7 @@ class Type(StrEnum):
         PARTIAL_TRANSLATED_TRANSCRIPTION = "partial_translated_transcription"
         PIPELINE_TIMINGS = "pipeline_timings"
         TTS_BUFFER_STATS = "tts_buffer_stats"
+        TTS_TEXT = "tts_text"
         ERROR = "error"  # For error messages
         END_TASK = "end_task"  # For end_task messages
         SET_TASK = "set_task"  # For set_task messages
@@ -131,6 +132,7 @@ class Type(StrEnum):
         Type.TRANSLATED_TRANSCRIPTION,
         Type.VALIDATED_TRANSCRIPTION,
         Type.PARTIAL_TRANSLATED_TRANSCRIPTION,
+        Type.TTS_TEXT,
     }
 
     IN_PROCESS_TYPES: ClassVar[set[Type]] = TRANSCRIPTION_TYPES
@@ -586,7 +588,7 @@ class TranscriptionMessage(Message):
     id_: str = Field(alias="transcription_id")
     text: str
     language: Language
-    segments: list[TranscriptionSegment]
+    segments: list[TranscriptionSegment] = Field(default_factory=list)
 
     model_config = ConfigDict(populate_by_name=True)
 
@@ -604,13 +606,15 @@ def extract_from_nested(cls, values: dict[str, Any]) -> dict[str, Any]:
             transcription = values["data"]["transcription"]
             # Convert language string to Language object
             lang_code = transcription["language"]
-            return {
+            result = {
                 "message_type": values["message_type"],
                 "transcription_id": transcription["transcription_id"],
                 "language": Language.get_or_create(lang_code),
-                "segments": transcription["segments"],
                 "text": transcription["text"],
             }
+            if values["message_type"] != "tts_text":
+                result["segments"] = transcription["segments"]
+            return result
         return values
 
     def model_dump(self, **kwargs) -> dict[str, Any]:
diff --git a/tests/test_enum.py b/tests/test_enum.py
index e7bedbe..e1a3e4d 100644
--- a/tests/test_enum.py
+++ b/tests/test_enum.py
@@ -32,6 +32,7 @@ def test_transcription_message_types():
         "translated_transcription",
         "validated_transcription",
         "partial_translated_transcription",
+        "tts_text",
     }
     assert TRANSCRIPTION_MESSAGE_TYPES == expected
 
@@ -44,6 +45,7 @@ def test_allowed_message_types():
         "translated_transcription",
         "validated_transcription",
         "partial_translated_transcription",
+        "tts_text",
     }
     assert ALLOWED_MESSAGE_TYPES == expected