4096 chunk sizes seams to be more robust

debug_000.py

@@ -16,7 +16,7 @@ def stream_tts(prompt):
             text=prompt,
             voice=voice_id,
             model=model_id,
-            stream_chunk_size=2048,
+            stream_chunk_size=4096,
             stream=True,
             )
         return audio_stream