xAI is rolling out 21 new flagship voices for Grok Voice, expanding its built-in voice roster from five to 26 and enhancing its capabilities in multilingual voice agents, text-to-speech, and no-code agent creation. These new voices are now available across the real-time Voice Agent API, Text-to-Speech API, and Grok Voice Agent Builder, targeting developers, operators, support teams, sales teams, educators, advertisers, and businesses building production voice agents.
Grok Voice now has 25+ flagship voices that each natively support 25+ languages! https://t.co/0QDsDEmsO3 pic.twitter.com/fNcmCd7bvk
— Eden Chan (@edenchan) July 7, 2026
The release introduces voices such as Carina, Zagan, Helix, Orion, Luna, Iris, Altair, Zenith, Perseus, Helios, Lux, Kepler, Rigel, Cosmo, Celeste, Ursa, Sirius, Lumen, Castor, Naksh, and Atlas. xAI positions them around specific use cases, including:
- Support
- Characters
- Commentary
- Advertising
- Education
- Wellness
- Narration
- Sales
- Assistants
- Podcasts
- Audiobooks
The original five voices, Ara, Eve, Leo, Rex, and Sal, have also been retrained for more natural pacing, phrasing, and emphasis.
The voice system operates across Grok’s real-time and TTS stack. Developers can stream audio and text over WebSocket for voice assistants, phone agents, and voice systems. Session settings support built-in or custom voices, server-side voice activity detection, tools, web search, X search, file search, MCP, function calls, pronunciation replacements, and adjustable speech speed. The current flagship real-time model is Grok Voice Think Fast 1.0, while the legacy Grok Voice Fast 1.0 model is marked as deprecated.
For text-to-speech, xAI supports expressive speech tags such as pauses, laughter, breathing, whispering, volume, pitch, and speed changes, singing, and emphasis. The TTS API accepts up to 15,000 characters, supports MP3 by default at 24 kHz and 128 kbps, and can return character-level timing metadata. Official documentation lists 20 supported TTS languages, while xAI’s announcement states the new voices are natively multilingual across Grok Voice’s 25+ languages.
The launch is closely tied to xAI’s new Voice Agent Builder, a beta no-code tool that lets users create voice agents in about two minutes. The builder includes features such as telephony, knowledge retrieval, tools, guardrails, MCP support, observability, SIP number support, call recordings, transcripts, and tool-use logs. xAI states that agents are billed at the API rate of $0.05 per minute of audio, with voice included and no separate platform fee, while telephony on a free-provisioned number costs an additional $0.01 per minute.
The company behind the release is xAI, which focuses on frontier reasoning, real-time voice, and generative media. The voice rollout expands Grok from a chatbot and API model family into a broader production stack for call centers, voice assistants, and brand-controlled speech experiences. Custom Voices are also significant: xAI allows users to clone a voice from a reference clip up to 120 seconds long and use it across TTS and real-time voice APIs. However, Custom Voices are currently available only in the United States except Illinois, with API creation gated to Enterprise teams.
