Transcription Delay
Time (seconds) between the end of speech and when the final transcript text is available.
LiveKit Agent Test Bench
Select a session preset, request a token, and connect to the agent room.
Select a realtime ElevenLabs agent or leave blank to use the built-in pipeline.
You can provide raw text or JSON; it will be forwarded to the token service.
Leave blank to use the URL returned by the token service (when available).
Audio Monitor
Status
Transcripts
User
Agent
Latency Metrics Explained
End of Utterance
Time from the VAD-detected end of speech until the user turn is considered complete. This already includes any transcription_delay.
TTS Duration
The amount of time (seconds) it took for the TTS model to synthesize the entire audio stream.
TTS Time To First Byte
Time (seconds) for the TTS model to produce the first byte of audio.
LLM Time To First Token
Time (seconds) for the LLM to emit the first token of the completion.
LLM Duration
The amount of time (seconds) it took for the LLM to stream the entire completion.
Total Latency
Total latency combines perception, reasoning, and speech: eou.end_of_utterance_delay + llm_metric.ttft + tts_metric.ttfb.