Voice

• List voices — browse TTS voices across all configured providers
• List models — discover TTS, STT, and STS models
• Create voice session — establish a real-time voice session over LiveKit WebRTC
• List providers — see which voice providers are enabled for your organization

​

Key concepts

• TTS (Text-to-Speech) — converts the agent’s text reply into audio streamed to the caller.
• STT (Speech-to-Text) — converts caller speech into text that the LLM processes.
• STS (Speech-to-Speech) — end-to-end speech model that bypasses the text layer entirely (lower latency, less control).
• Voice pipeline — the combination of STT → LLM → TTS that processes a voice turn.
• voicePipelineMode on agents:
  • pipeline — classic STT + LLM + TTS
  • sts — direct speech-to-speech
  • hybrid — adaptive mode that switches based on conditions

​

Connecting to a voice session

import { Room, RoomEvent } from 'livekit-client';

const { token, serverUrl } = await fetch('https://api.agntix.ai/v1/voice/sessions', {
  method: 'POST',
  headers: { 'x-api-key': 'pk_live_••••', 'Content-Type': 'application/json' },
  body: JSON.stringify({ agentId: 'a1b2c3d4-...' }),
}).then(r => r.json());

const room = new Room();
await room.connect(serverUrl, token);
room.on(RoomEvent.TrackSubscribed, (track) => {
  // play agent audio
});

​

Inbound calls