diff --git a/src/pipecat/services/deepgram.py b/src/pipecat/services/deepgram.py
index 3750164cf..eaa82a6c9 100644
--- a/src/pipecat/services/deepgram.py
+++ b/src/pipecat/services/deepgram.py
@@ -97,17 +97,16 @@ class DeepgramTTSService(TTSService):
 
             # Read and yield the audio data in chunks
             audio_buffer.seek(0)  # Ensure we're at the start of the buffer
-            chunk_size = 8192  # Use a fixed buffer size
+            chunk_size = 1024  # Use a fixed buffer size
             while True:
                 await self.stop_ttfb_metrics()
                 chunk = audio_buffer.read(chunk_size)
                 if not chunk:
+                    yield TTSStoppedFrame()
                     break
                 frame = TTSAudioRawFrame(audio=chunk, sample_rate=self.sample_rate, num_channels=1)
                 yield frame
 
-                yield TTSStoppedFrame()
-
         except Exception as e:
             logger.exception(f"{self} exception: {e}")
             yield ErrorFrame(f"Error getting audio: {str(e)}")