Grok Voice (xAI)
Configure Grok voice, tools, limits and estimated costs.
Select Grok Voice 2.0 (xAI) in the bot voice settings, then select an xAI voice (default: Eve). Speech speed supports 0.7–1.5. The provider uses server VAD with automatic responses and interruptions; semantic VAD and disabling automatic responses or interruptions are unsupported.
The existing BeAI skills, permissions and organisation boundaries apply. Tool calls execute through BeAI and return their results before the conversation continues. Native paid xAI search tools are not enabled automatically.
The model needs an active price in your organisation and its hierarchy before it can be selected. An administrator can initialize it through the usual pricing workflow.
Select a voice, then use the preview button beside the selector. All 28 xAI voices have samples in French, Belgian French, Canadian French, English, Dutch, Spanish and Catalan. Playback follows the bot language, with fallback to its primary language, the interface language and French. The same button supports Gemini voices. Samples use a fixed demonstration sentence at normal speed, without the bot's private instructions, greeting or customized speed. Previewing does not start a call or change the bot, and existing form permissions still apply. If no sample can be played, playback stops; try again or choose another voice.
The equivalent public operation is PATCH /api/v1/bots/{id}/voice:
{"ai_model":"grok-voice-think-fast-2.0","voice":"eve","speech_speed":1.0}
GUI and API reject incompatible voices, unsupported VAD options and speeds below 0.7. The xAI key remains server-side. Missing credentials, insufficient provider credits and concurrency limits may prevent a call from starting; inspect the call history and contact the platform administrator.
BeAI credits follow the configured model rate. Provider costs are BeAI estimates of transmitted and received audio seconds and client text messages, excluding tool results. These estimates are not xAI invoices. Published prices on September 16, 2026 are USD 0.08 per audio minute sent or received and USD 0.004 per text message. Per-metric supplier prices are configurable per million seconds or messages. The documented processing region is us-east-1, with 10 concurrent sessions per team and a 120-minute session limit.