Voice settings

For local testing, point at your backend, choose a provider, and pick a voice.

Voice provider

Azure Realtime: WebSocket directly to /voice. ElevenLabs: bootstraps via /api/voice/session?provider=elevenlabs then connects to the ElevenLabs signed URL; the SerenPath system prompt + crisis pipeline run via the Custom LLM webhook.

Backend WebSocket URL (Azure Realtime only) Backend HTTPS URL (for dev token + ElevenLabs session) Azure Realtime voice (multilingual) sage — warm, clinical (recommended for grief) marin — natural prosody, newer ballad — slower, thoughtful coral — warm, slightly upbeat alloy — neutral, slightly bright ash — alternate verse — expressive

All Azure voices are multilingual — they switch languages automatically based on what Whisper detects.

ElevenLabs voice (live from your account) ↻ refresh (use agent default) loading from ElevenLabs…

Pulled live from /api/voice/elevenlabs/voices (= your "My Voices" in the ElevenLabs dashboard). To add voices: browse Therapist, Soothing, Gentle, Sympathetic categories → "Add to My Voices" → hit refresh here. Override applies to next session.

User identity Show diagnostic log