Voice Latency Playground

Test voice pipeline latency: STT / LLM / TTS

disconnected

LLM Strategy

STT Provider / Model

Provider

Model

TTS Provider / Model

Provider

Model

LLM Provider / Model

Infrastructure

Provider

Model

Cache Preheat

Warm LLM cache on connect

Fires a background LLM call at connect time to prime the provider's KV cache, reducing first-turn TTFB.

Debug Frames

Log TTS frame events to server

Writes every TTS frame event (start, audio chunks, stop) to a JSONL log file on the server for debugging audio issues.

Conversation Stage

TRIAGEPRICINGNAME COLLECTIONADDRESS COLLECTIONAPPOINTMENT BOOKINGFAREWELLESCALATION

Conversation State

No state data yet. Connect and start talking.

Latency Per Turn

No latency data yet. Connect and start talking.

Conversation

Connect to start a voice conversation.