Baseline Voice Agent

Sequential STT → LLM → TTS pipeline with Google Gemini models

idle

Pipeline Stage Tracker

Visual execution breakdown of the sequential voice pipeline

Status: Idle

1. VADWaiting

Browser RMS

2. STTWaiting

gemini-3.5-transcribe

3. LLMWaiting

gemini-3.7-flash

4. TTSWaiting

gemini-3.1-flash-tts

5. AudioWaiting

Browser Playback

User Transcript (STT)

Waiting for user speech...

Agent Response (LLM)

Waiting for LLM generation...

Turn Latency Measurements

Empirical benchmarks including LLM Time-To-First-Token (TTFT)

Total Latency: p50: 0ms | p95: 0ms

LLM TTFT: p50: 0ms | p95: 0ms

Turn	VAD (ms)	STT (ms)	LLM TTFT	LLM Total	TTS (ms)	Total (ms)
No turns recorded yet. Click 'Start Speaking' to run an interaction.

Pipeline Logs Console

Ready. Start an interaction to stream pipeline logs...