Introduction
ProsodyAI — live tone, speaker turns, and agent steering for voice AI
ProsodyAI sits beside your speech-to-text stack and reads how a call is going — stress, engagement, turn-taking, overlap — so your agent can adapt mid-call and your team can review the sessions that matter.
Start here
Quickstart
Analyze a recording with the TypeScript SDK
Integration
Batch REST vs live WebSocket streaming
TypeScript SDK
Client methods and response types
Local demo
Run the sidecar and hear tone change live
What you get back
| Output | When |
|---|---|
| Valence / arousal / dominance | Batch and streaming |
| Delivery signals (engagement, stress, certainty, …) | Batch and streaming |
| Per-turn prosody (interview / call review) | Batch with diarization |
| Transcript + speaker labels | Batch and streaming |
| Agent steering cues | Streaming — when tone shifts |
| Org KPIs | When configured for your organization |
Auth
Create a psk_* API key in the dashboard under Organization → API keys. Use the same key for batch and streaming.