Build
Choose a voice model
Compare the four managed voice choices and their deployment requirements.
Managed choices
In an agent's Voice model selector, choose the engine that fits the conversation. The displayed retail rate is for AI voice usage; carrier charges and number rental are separate.
| Voice model | Intended fit | AI voice rate |
|---|---|---|
| Mistral | Everyday calls with Voxtral transcription and speech plus Mistral reasoning. | $0.15/min |
| Gemini Live | A balanced live-audio option. | $0.20/min |
| GPT Realtime 2.1 Mini | A lower-cost GPT live-audio option. | $0.25/min |
| GPT Realtime 2.1 Standard | The higher-priced GPT live-audio option. | $0.50/min |
Voice minutes are billed by the second under the account's active rate card. Check Billing for the applicable account configuration. These are customer prices, not a provider-cost estimate or a quality guarantee.
Select and test
- Select the model in the agent editor.
- Choose one of that model's offered voices and preview it.
- Save the draft and start a fresh audio test.
- Compare the same scenario, language and tool path before choosing.
Voice catalogs differ. Changing engines can reset an incompatible voice. An Assistant chat model is a separate choice and does not switch the agent's call engine.
Availability depends on the deployment. The native SIP gateway currently uses the Mistral pipeline; choosing another model in the editor does not provision it on that gateway. Check the deployment's supported engine before routing a clinic number. Automated simulation suites currently support the Mistral pipeline; other managed realtime engines can be checked with individual editor tests.
Provider credentials for managed engines are operated by Verbatik. Your clinic supplies its own credentials for external booking, CRM and knowledge integrations.