I'm glad you reached out! LiveKit is designed to process and deliver transcriptions for both the agent and the user independently, even when they’re speaking at the same time. Each transcription segment comes with helpful timing metadata like

firstReceivedTime

and

lastReceivedTime

, which makes it easier for your frontend applications to manage and display overlapping speech smoothly. To tackle the issue you're facing, you might want to add some logic in your frontend that sorts and displays transcriptions based on their timestamps. Also, if you’re using ElevenLabs or Cartesia TTS providers, enabling TTS-aligned transcription forwarding can really help because it provides word-level timing data for perfect syncing of text with speech. You can set this up by adding

use_tts_aligned_transcript=True

in your

AgentSession

configuration. If you’d like, I’m here to help you step-by-step with implementing these solutions—just let me know! Sources: ElevenLabs TTS integration guide | LiveKit Docs | Understanding Transcript Timing in Voice Conversations | AI voice agents | LiveKit Docs | complex-agents/nutrition-assistant/nutrition-assistant-frontend/hooks/useCo... | Docs::Agents