Thanks — I hear you: you’re seeing the same spoken sentence printed multiple times in your UI (e.g., “hi hello” shown three times in a room with three participants). I’ll help you narrow the cause and give concrete fixes. First, quick diagnosis (what’s most likely happening) • Each participant or agent in the room is publishing the same transcription text (or an interim + final version) and your frontend is rendering every incoming text event without deduplication. • Transcriptions include a

segment_id

and an attribute

lk.transcription_final

. The same segment can be emitted by multiple sources or as interim + final updates, which will look like duplicates unless you filter them. What you can do right now (recommended fixes) 1. Use

segment_id

to dedupe and prefer final segments ◦ Only render when

lk.transcription_final === "true"

(ignore interim updates if you want only final text). ◦ Track seen

segment_id

values (Set or map) and skip duplicates. ◦ Example approach: ▪︎ On text stream event, read

segment_id

and

lk.transcription_final

. ▪︎ If not final, ignore (or show as temporary UI). ▪︎ If final and

segment_id

already processed, skip. 1. Filter by sender identity when appropriate ◦ If only one participant/agent should publish transcriptions, check

participantInfo.identity

and only render from that identity. 1. Ensure a single source generates transcriptions ◦ Audit your setup: do you have client-side STT (Web Speech API) + an agent STT + relays? If multiple systems transcribe the same audio, disable redundant paths so only one publishes to the room. 1. Use the React helper if available ◦ If you use

@livekit/components-react

, the

useTranscriptions()

hook handles deduplication and final/interim handling for you. Consider switching to that hook to simplify logic. Short sample logic (conceptual) • Maintain a Set of processed segment IDs. • On incoming transcription event: a. Read

segment_id

and

lk.transcription_final

. b. If not final -> ignore or buffer. c. If final and not in Set -> render and add to Set. d. If final and in Set -> ignore. A couple of quick questions so I can tailor exact code to your setup • Are you using

@livekit/components-react

and

useTranscriptions()

or a custom text stream handler? • Do you have any agent(s) or client-side STT running simultaneously (server agent + client speech capture)? • Can you share a small snippet of how you currently handle incoming transcription events in React? If you share those details I’ll give a short, copy-pasteable React example that matches your setup and fixes the duplicate output. Sources: Text and transcriptions | LiveKit Documentation | How to build an agent with speech-to-text input and text-only output | Live... | Pi Zero Transcriber | LiveKit Documentation