SAN FRANCISCO — xAI has expanded Grok's voice lineup with 21 new flagship voices and a dedicated Voice Agent Builder, a release that turns Grok from a chatbot you type to into a platform for building AI agents that listen and speak in real time.
The new voices are all multilingual and available across xAI's realtime Voice Agent API and its Text-to-Speech API. Paired with the Voice Agent Builder, the update gives developers a low-friction way to assemble customer-service lines, tutors, in-car assistants and other spoken applications on top of the Grok 4.5 model.
What Shipped
The headline is breadth. Twenty-one distinct voices, each able to switch languages mid-conversation, gives builders a palette that used to require stitching together several vendors. xAI also added the ability to clone a voice from a short recording and deploy it across both the streaming Voice Agent and the standard Text-to-Speech endpoints.
The Voice Agent Builder sits on top of all of it, letting developers wire a spoken persona to Grok's reasoning without managing the plumbing of speech recognition, turn-taking and synthesis themselves. It is the natural extension of the automation push we covered when Grok added scheduled, hands-off tasks to its consumer app.
Built on the Grok 4.5 Engine
The voice tools run on Grok 4.5, xAI's most capable model for coding, agentic work and knowledge tasks. That matters because a voice agent is only as good as the brain behind it — natural speech falls apart the moment the underlying model gives a wrong or slow answer. By anchoring the voices to 4.5, xAI is betting that response quality, not just audio fidelity, is what will win developers.
The company has been shipping at a rapid clip. The same developer surface recently gained parallel task execution, as we detailed in our look at Grok Build's workflows that fan tasks across many agents. Voice is the next layer on that stack, and xAI is offering free Grok 4.5 usage in Grok Build for a limited time to pull builders in.
Why It Matters
Voice is shaping up to be the interface where AI assistants feel most human, and it is also the hardest to get right. By bundling a large voice library, cloning, real-time APIs and a no-plumbing builder, xAI is trying to make Grok the default choice for anyone shipping a talking agent. Full release notes are posted on xAI's news page.
Expect the voice push to feed directly into Musk's broader ambitions, from in-car Grok assistants to the conversational layer for humanoid robots, as xAI keeps folding its models into the wider Tesla and SpaceX ecosystem.
