xAI has launched a new feature called "Custom Voices" that enables users to clone their own voice from a short recording. The process requires only about a minute of natural speech captured through the xAI console. According to xAI, the voice model is ready in under two minutes and can be integrated into the company's text-to-speech and voice agent APIs.
To prevent misuse, xAI implements a two-step verification process. Users first read a passphrase that is verified in real time, and then the system compares voice characteristics of both recordings to ensure the same person is speaking. xAI claims this setup makes it impossible to clone existing recordings or another person's voice.
The xAI console also introduces a new "Voice Library" featuring more than 80 preinstalled voices across 28 languages. Using cloned voices comes at no additional cost.
Custom Voices builds on xAI's recently launched Grok Speech-to-Text and Text-to-Speech APIs and the Grok Voice Think Fast 1.0 voice agent model. xAI states that this model already powers Starlink's customer support and sales operations.