does using an external stt remove the users turns ...
# ask-aib
bulky-actor-93512
09/23/2025, 3:02 AMdoes using an external stt remove the users turns frfom the transcrippts dict?
t
tall-airline-16756
09/23/2025, 3:02 AMThanks for reaching out, Sebastian. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.
tall-airline-16756
09/23/2025, 3:03 AMUsing an external STT doesn’t automatically remove user turns from the transcripts dictionary. The transcripts get cleared based on end-of-turn (EOT) detection, which depends on how your turn detection is set up (like VAD, STT, or context-aware models) and what the external STT supports. If your external STT can emit
END_OF_SPEECHturn_detection="stt"b
bulky-actor-93512
09/23/2025, 3:04 AMsession = AgentSession(
stt=deepgram.STT(model="nova-3", language="en-US"),
llm=openai.realtime.RealtimeModel(
model="gpt-realtime",
turn_detection=None,
input_audio_transcription=None
voice=config.voice,
temperature=config.temperature
),
vad=ctx.proc.userdata["vad"],
turn_detection=EnglishModel(),
)
with this setup i dont have the users turns in the transript
t
tall-airline-16756
09/23/2025, 3:04 AMThe issue with missing user turns in your transcript might be due to a little hiccup between the OpenAI RealtimeModel's built-in turn detection and LiveKit's turn detection model. OpenAI's Realtime API handles speech-to-text internally and only sends transcripts after a turn is finished, which can sometimes clash with LiveKit's turn detection that works best with live, incremental STT results. Here’s a friendly checklist to help you resolve this: 1. Try disabling OpenAI's turn detection by setting
turn_detection=NoneEnglishModelturn_detection