Use the following guides to configure the Nemotron Voice Agent. Each guide is task-focused and includes the steps to enable or configure specific features.
| Guide | Description |
|---|---|
| Switch LLM Models | Change the LLM backend between local NIM microservices and NVIDIA cloud endpoints. |
| Customize System Prompts | Select pre-built prompt samples or create custom prompts to define your agent's personality and behavior. |
| Enable Multilingual Voice Agent | Deploy the agent with automatic language detection and multilingual responses. |
| Configure TTS Settings | Set up TTS voice, pronunciation correction (IPA), cloud TTS endpoints, and text filters. |
| Enable Zero-Shot TTS | Clone any voice from a short audio sample using the Magpie Zero-shot model. |
| Choose a Transport Method | Switch between WebRTC (default) and WebSocket transport. |
| Enable OpenTelemetry Tracing | Add observability with Phoenix to monitor pipeline performance and conversation flows. |
| Tune Pipeline Performance | Adjust speculative speech, chat history, audio debugging, and output buffering settings. |