Xai Custom Voices: custom voices for voice agents via API
Xai introduced Custom Voices and Voice Library on April 30, 2026. A custom voice can be created from a short recording and used in the TTS and Voice Agent APIs.
This is not a new frontier model. But it is a practical release for companies taking voice agents seriously. Voice is no longer just a technical output. It is becoming part of brand, customer experience and governance.
What changes
Xai lets teams create a custom voice from reference audio. The documentation mentions up to 120 seconds of audio and up to 30 custom voices per team. API-based voice creation is available for Enterprise teams.
Custom voices can be used through TTS APIs and Voice Agent APIs over REST or WebSocket.
Xai also describes controls based on a passphrase and speaker similarity, intended to make unauthorized voice cloning harder.
What it means in practice
For companies, the product angle is the most interesting part. If you build voice support, internal voice assistants or automated audio output, a custom voice can be the difference between a technical demo and something that feels consistent.
Relevant scenarios include customer support, onboarding, product tutorials, multilingual voice workflows and voice-agent prototypes without a recording studio.
Where to wait
Voice cloning is useful, but sensitive. I would not use it in customer communication without internal rules.
At minimum, you need explicit consent from the person whose voice is used, an audit trail, retention rules for source recordings and an incident process if the voice is misused.
Conclusion
Xai Custom Voices will not change the AI stack by itself. But it shows voice agents moving from demo to product layer.
If your company only uses AI as a text chatbot, this may not matter yet. If you work on support, onboarding, call centers or internal assistants, custom voices will soon be a normal requirement.