Hi! I have a question regarding the creation of agents using OpenAI’s Realtime LLM in text mode together with ElevenLabs TTS. I’m not able to generate the full audio — the generation stops halfway, and the audio gets cut off before finishing.

Are you on the latest version of agents? And configuring your session as follows: https://docs.livekit.io/agents/models/realtime/#separate-tts ?

Hi, yes I am, here's the code I'm currently using. My problem is only happening when using the Elevenlabs plugin; it works normally with Azure Speech. session = _AgentSession_( userdata=userdata, llm=openai.realtime.RealtimeModel._with_azure_( modalities=["text"], ), tts=elevenlabs._TTS_( _voice_id_="cFylwQo5ufGYUNyRS167", ), )

😕 I can't match your setup because I don't have a test key for Realtime OpenAI with Azure so I'll have to ask around or create one. I did try with OpenAI realtime + Elevenlabs TTS and that config does not show any cut off. Do you see anything in the agent observability view? https://docs.livekit.io/agents/observability/, that should give some indication if the TTS is being cut short.