Thanks — this is a clear question. You’ve already configured

stt

in your

AgentSession

(good first step). The reason you’re seeing transcriptions of the agent but not the user is that user speech transcriptions are emitted on a different event:

user_input_transcribed

. To receive what the user says you need to subscribe to that event on your

AgentSession

. What to do next 1. Add an event handler for

user_input_transcribed

on your

AgentSession

. 2. Inspect

transcript.alternatives[0].text

and

transcript.is_final

to get final vs interim text. Python example

@session.on("user_input_transcribed")
def on_user_transcription(transcript):
    if transcript.is_final:
        print("User said (final):", transcript.alternatives[0].text)
    else:
        print("User said (interim):", transcript.alternatives[0].text)

Quick checklist if you still don’t see anything • Confirm the user’s microphone audio is being sent from the client to the room. • Confirm the agent/session is connected and subscribed to the user’s audio track(s). • Verify you aren’t filtering out

user_input_transcribed

events elsewhere in your code. • STT is already configured correctly in your snippet — just ensure the live audio stream is available. If you’d like, share: • Whether audio is working for other features (e.g., live audio in the room), • Any logs showing event names being received, and I’ll help narrow down why

user_input_transcribed

might not be firing. Sources: Speech-to-text (STT) models overview | LiveKit Documentation | Transcriber | LiveKit Documentation