xAI introduces voice cloning on Grok and structures its catalog with a Voice Library

xAI launches Custom Voices on Grok, a tool cloning voices in under two minutes for its TTS and Voice Agent APIs, managed via a new 80-voice Voice Library.

xAI has released the Custom Voices feature for Grok, which allows cloning a voice from approximately one minute of audio recording and using it in the publisher's Text to Speech and Voice Agent APIs. The pipeline delivers a production-ready voice model in less than two minutes. Once generated, the voice ID can be passed to any TTS endpoint or used with the Voice Agent API for real-time conversational agents. The cloned voice inherits all TTS capabilities, including speech tags, multilingual output, and REST or WebSocket streaming.

Cloning is subject to a two-step verification mechanism designed to prevent fraudulent use. The user first reads a validation phrase aloud, which is transcribed and matched live by xAI's Speech-to-Text engine to confirm presence and consent. The system then calculates speaker embeddings on the validation phrase and the full recording, and verifies that they correspond to the same person. Cloning from a pre-existing recording or a third-party voice is technically blocked.

xAI is also releasing a Voice Library, a console-integrated management space that brings together custom voices and built-in voices. The publisher's own catalog has been expanded to include over 80 voices across 28 languages, which can be previewed in different scenarios before integration. The use of Custom Voices with the TTS and Voice Agent APIs does not incur any additional cost.