Each new voice is natively multilingual across 25+ languages and cast for a specific role — support, characters, commentary, advertising, education — and the original five voices were retrained for more natural delivery.
2026-07-13 · by Moe Ameen
xAI expanded Grok Voices in early July 2026, adding 21 new flagship voices to the five it already offered and bringing the roster to 26. The company's framing is naturalness and range: rather than one generic narrator, each voice was cast for a specific job — support agents, character voices, commentary, advertising reads, and education. The new set carries names like Carina, Orion, Luna, Iris, Altair, Zenith, Perseus, Helios, Kepler, Rigel, Cosmo, Celeste, Sirius, and Atlas.
Every flagship voice is natively multilingual, supporting all of Grok Voice's 25+ languages, so a single voice keeps its character from one language to the next. The original five voices — Ara, Eve, Leo, Rex, and Sal — were retrained in the same update for more natural pacing, phrasing, and emphasis. Delivery is steerable through speech tags for pauses, whispering, laughter, and changes in volume, pitch, speed, and emphasis.
The voices are available through xAI's realtime Voice Agent API, its Text-to-Speech API, and a new Grok Voice Agent Builder for assembling agents without wiring the infrastructure yourself. xAI also offers voice cloning from a reference clip up to 120 seconds, usable across the TTS and realtime APIs, currently limited to the United States. Exact pricing and model details move with each release, so confirm current numbers in xAI's own docs.
A Grok voiceover is a starting point, not a post. The voice engine outputs audio and stops — no picture, no captions, no feed. Kompozy is a content generation and publishing engine, not a voice model, so the two slot together instead of competing. Take a script voiced in one of the new flagship voices — an education voice for a how-to, a commentary voice for a hot take — and let Kompozy build the video around it: a Persona Short or HeyGen avatar clip with a face-locked recurring identity, a Listicle Video, or a Clipped Short from longer footage. Kompozy then burns in word-synced captions, reframes to 9:16, 1:1, or 16:9 per feed, and schedules and publishes the finished set across nine social platforms plus blog and email from one queue with Autopilot and a per-post review pipeline.
There's also a same-week play. "xAI just shipped 26 natural, multilingual Grok voices" is a topic your audience is reading right now. Drop your take into Kompozy Quick Ingest and one point of view fans out into a blog explainer, a captioned short, a brand-exact Carousel through HyperFrames, and platform-native Text Posts — all held to your voice by the Persona Brief — so you publish a clear angle on the news across every surface while it's fresh. Better, cheaper voice generation only pays off when something turns the voice into finished content; Kompozy is that something.
Twenty-six. xAI added 21 new flagship voices in early July 2026 to the original five (Ara, Eve, Leo, Rex, Sal), which were retrained in the same update for more natural pacing, phrasing, and emphasis. Each voice is natively multilingual across 25+ languages.
Through xAI's APIs — the realtime Voice Agent API, the Text-to-Speech API, and a new Grok Voice Agent Builder. There is also voice cloning from a reference clip up to 120 seconds, usable across the TTS and realtime APIs, currently limited to the United States.
No. Grok Voices generates the audio layer — narration, real-time voice agents, and cloned voices — but produces no captioned video, carousels, blogs, or scheduled posts. To turn a Grok voiceover into finished, on-brand content across platforms, creators pair it with a content engine like Kompozy that generates the video, adds captions, and publishes it.