Thanks — I’m responding to your report that the LiveKit Playground isn’t showing transcriptions in the chat while you’re testing with OpenAI Realtime models on livekit-agents v1.3.4. I can help troubleshoot — a few likely causes and quick checks below, plus one question to narrow things down. Possible reasons and quick checks • Realtime models don’t emit interim transcripts and user transcriptions can arrive delayed. Check if transcripts appear after the agent reply (delay is normal). • If audio output is disabled, the playground’s legacy UI may not show text-only transcripts. Verify the agent publishes text output and/or audio: ◦ Ensure

text_output=True

on the agent. ◦ If you have audio disabled intentionally, confirm you’ve implemented a custom receiver for the

lk.transcription

text stream (the playground’s default UI may not display these). • Transcriptions for agent speech are published to the

lk.transcription

text stream topic. If that stream isn’t being forwarded to the frontend, the chat won’t show them. • You may need a custom text-stream handler or transcription_node adjustments if formatting or sync is required. One clarifying question • Do you see any transcription messages at all (e.g., in logs or the Cloud Dashboard transcripts), or is nothing present server-side? If you can share: • Whether audio output is enabled or disabled, • Any relevant logs or messages that show transcription publishing (or nothing), • Whether transcripts appear after a short delay, I’ll give the exact next steps to check the

lk.transcription

stream, confirm

text_output

settings, and (if needed) show how to add a custom text-stream handler or enable audio output so the playground UI shows transcripts. Sources: Text and transcriptions | LiveKit docs | Realtime model plugins | LiveKit Docs | How to Set Up Speech-to-Text Input with Text-Only Output in LiveKit Agents | Pipeline nodes and hooks | LiveKit docs | Streamline Troubleshooting with Agent Observability