Voice Agent Prototypes

Setup

Scenario

Prototype A: STT → LLM → TTS B: speech-to-speech Blind (random A/B)

Options Patient input End-of-turn silence (ms) Allow barge-in (A) Phone-line quality (8 kHz μ-law) Live captions on phone Send patient name to LLM (A) Record audio (participant consented) Tester name

Call request JSON (PRD input)

Ready

Health Complex

+880 ···

00:00

Calls

When	Engine	Scenario	Status	Outcome	Conf.	Secs	Latency	Cost	Rated

Calls

Prototype comparison

Access code