xAI Custom Voices: Clone Any Voice in Under 2 Minutes

xAI shipped Custom Voices on April 30, 2026, adding voice cloning to its Grok API platform. Users record about 60 seconds of natural speech through the xAI console, and the system returns a production-ready voice model in under two minutes. The cloned voice works across Grok’s Text-to-Speech and Voice Agent APIs at standard API rates. xAI also expanded its built-in voice catalog to over 80 options in 28 languages.

At $3/hour for voice agents, xAI is undercutting ElevenLabs and OpenAI on price by a wide margin. The feature set is thinner, but the economics change the math for anyone building voice into a product.

How Does xAI Voice Cloning Work?

The cloning process runs entirely through the xAI console. Users read aloud several passages of unrelated dialog while the system records. A two-stage verification pipeline handles the rest: first, the speaker reads a verification phrase that Grok’s speech-to-text engine transcribes and matches in real time, confirming intent and presence. Then the system computes speaker embeddings from the verification clip and the full recording to confirm both belong to the same person.

This design means you cannot clone a voice from a pre-existing audio file, and you cannot clone someone else’s voice. Once verified, the system processes the recording and delivers an 8-character alphanumeric voice_id that works wherever xAI’s built-in voices do. Each team can create up to 30 custom voices simultaneously, and any voice can be deleted with a single click.

What Is the Voice Library?

The Voice Library is a new section in the xAI console that puts all available voices in one place. Custom voices show up next to the five built-in options (Eve, Ara, Rex, Sal, and Leo). With this launch, xAI also expanded the pre-built catalog to over 80 voices in 28 languages. You can preview any voice across different scenarios before picking one.

Each built-in voice has a different personality: Eve is energetic, Ara is warm and conversational, Rex leans professional, Sal is smooth, and Leo sounds authoritative. Custom voices get the same TTS capabilities as built-ins, including inline speech tags for whispers, laughter, sighs, and emphasis. Output works over both REST and WebSocket streaming.

How Much Does xAI Voice Cloning Cost?

There is no extra charge for using custom voices. The pricing follows standard xAI API rates:

The Voice Agent API runs on grok-voice-think-fast-1.0, which combines reasoning with real time speech. It supports tool use — web search, X search, file search, and external MCP server connections — so the agent can actually do things mid-conversation, not just talk. For client side applications, Ephemeral Tokens let you open WebSocket connections without exposing your main API key.

Programmatic access to the custom voice creation endpoint (POST /v1/custom-voices) is currently limited to teams on an Enterprise plan. The console-based voice creation tool is open to all users with API access.

How Does xAI Voice Pricing Compare to ElevenLabs?

The pricing difference between xAI and ElevenLabs is large, though they’re not selling the exact same thing:

ElevenLabs still has the bigger voice library, works everywhere, and runs the Iconic Marketplace for licensed celebrity voices. xAI wins on voice agent pricing and doesn’t charge for custom voice creation. ElevenLabs requires at least a Starter subscription ($5/month) before you can clone anything.

If you’re outside the US or need access to a bigger voice catalog, ElevenLabs works globally and has 3,000+ voices available today.

What Safety Measures Are in Place?

xAI’s two-stage verification is stricter than what most voice cloning platforms require. The real time phrase matching confirms the speaker is physically present during the cloning session, not submitting a pre-recorded file. The embedding comparison then checks that the verification phrase and the full recording actually come from the same person.

Custom voices stay private to the team that created them. xAI says audio data is processed in real time and never stored or used for training. The platform has SOC 2 Type II certification, HIPAA eligibility, and GDPR compliance for European data — though the cloning feature itself is still US only.

What This Means

For Voice AI Developers

$3/hour voice agents change the economics for anyone running voice at volume. Customer support bots and IVR systems that cost $10-18/hour on ElevenLabs suddenly make more sense on xAI’s stack. The OpenAI Realtime API compatibility also means existing voice apps built for OpenAI can switch over without rewriting much code.

For the Voice Cloning Market

Voice cloning now has three tiers. ElevenLabs has the most features, the biggest library, and global reach — we cover the full landscape in our best AI voice generators roundup. OpenAI sits in the middle with TTS in ChatGPT. xAI is the cheapest option by far, with stricter verification than either competitor.

The US-only restriction matters a lot. Anyone outside the States still can’t create custom voices, which keeps ElevenLabs as the default internationally. For free alternatives, see our best free voice cloning tools guide. If xAI opens this up to more countries, the pricing pressure on everyone else gets real.

FAQ

---

Sources

1. Custom Voices and Voice Library - xAI Official - April 30, 2026
2. xAI launches Grok 4.3 and voice cloning suite - VentureBeat - May 1, 2026
3. xAI’s Custom Voices feature - The Decoder - May 2, 2026
4. xAI Voice API Documentation - May 2026