| Timestamp | Verdict | Confidence |
|---|
| Speaker | Emotion pattern | Speaking time | Language |
|---|
| Speaker | Behavior | Model reasoning | Confidence |
|---|
| Timestamp | Music | Speech |
|---|
Use the endpoint's built-in default, or build a custom configuration (conversation types, roles, behaviors, model signals). The right pane shows exactly what's sent as config to /api/velma-2-batch (batch) or as the first frame on /api/velma-2-streaming (streaming).
The pane on the right is the exact config that will be sent. It stays the literal "default" — the server's built-in configuration — until you add or change something below, at which point it becomes an explicit BatchConfig. Reset returns it to "default".
Candidate types the model picks one of (or none). The endpoint has no built-in catalog — you define them.
Candidate roles the model assigns to each speaker. You define them.
Pick server presets (added as preset:<id> refs the server expands), or define your own. Each becomes an entry in behaviors.
Start building with Modulate's voice intelligence APIs.