The speaker is the state.

A voice agent that recognizes who is speaking, resumes each person’s conversation across calls, and listens while it speaks.

Try the API

Model

ProsodySSM

A recurrent state-space model whose state is the speaker. Audio arrives continuously and every 80ms frame steps the state forward, one fixed-size state per person, held across the whole call and across every call after it. Identity, continuity, presence, and significance are four readouts of that one trajectory: who you are, the conversation so far folded into a state, where you are right now relative to yourself, and what this moment moved in you.

Product

API

One model listens to the conversation and returns its readouts: the transcript, who is speaking, and how each person is moving against their own baseline. Build with the TypeScript SDK, the LiveKit agent plugin, and the dashboard.

Conversation continuity for enrolled people: one state resumed across calls
Realtime diarization with unlimited speakers
Prosody analysis of pitch, loudness, and pacing per speaker
Speaker-relative change, measured against each person’s own baseline

Try the API Docs

Enterprise

An agent that listens while it speaks and picks up where the last call ended. Callers your organization enrolls, with their consent, never repeat themselves.

Saved conversation context that outlives every call, for people enrolled with consent
Your KPIs, declared as a schema and scored on every finished call
Moments replayable from their captured state
Conversation intelligence on every call

Get in touch

Contact

Talk to our team

Get a demo of the API and the full-duplex agent stack
Discuss volume pricing and enterprise plans
Review security, privacy, and data retention
Plan your integration with our engineers