Gemini now speaks with 66 regional voices, from U.K. English to Korean
Google says the 66 new voices carry local rhythm, pitch and inflection across Gemini Live, chat read-aloud and language practice, and work on Android and iOS.

Google is rolling out 66 regional voices in the Gemini app on Android and iOS, according to an official post from Neel Joshi, director of product management for Gemini Live. The voices are tuned to local speech patterns, not to caricature accents.
Seven named languages, one voice across the app
The post names seven examples: English as spoken in the U.S., the U.K. and India, plus Hindi, Brazilian Portuguese, Indonesian and Korean. Google says each of them gets tailored options. It says the 66 voices include these, and the wording leaves open whether other languages or dialects are covered. Google does not publish a full list or a per-language count in the post.

Google describes the voices as tuned for natural cadences and everyday clarity. They are meant to capture familiar local rhythm, pitch and inflection. The company says it is deliberately avoiding exaggerated accents, which is a design choice as much as a technical one. A voice that sounds like a parody of a region is easy to build, and a voice that sounds ordinary to locals is harder.
The chosen voice follows the user everywhere in the app. Google lists Gemini Live conversations, language practice, and the Listen icon in chat, which reads a long response aloud. To try them, tap the three-dot menu in Gemini Live, or open Gemini’s voice in the app settings. Both routes let users preview the options and pick one.

The post gives no plan requirements, price or country list. Google says only that the rollout is under way on Android and iOS. Whether every voice reaches every account at once, and whether the feature depends on the user’s language setting, is not stated.
The feature was first promised earlier this year. Joshi writes that at Google I/O the company shared plans to bring local dialects to Gemini so people can choose a voice that resonates with them. Today’s post is the delivery on that plan.
Sixty-six voices against a much larger voice catalog
Sixty-six is a sizeable number for a consumer app, but it sits next to much larger counts on Google’s developer side. Our earlier report on Gemini TTS leaving preview described Gemini 3.8 Flash TTS as covering 130 languages, with Flash-Lite TTS covering 101 in a comparison table. That coverage came from a Gemini API changelog entry dated September 22, 2026. The same report described an Extended Voice Library of more than 150 prebuilt and custom voices.
Set against those figures, the app release is narrower. Seven named languages are a small slice of 130, and 66 voices are fewer than half of 150. The two products are different, though. The API numbers describe what developers can generate. The app numbers describe what a consumer picks from a menu. The Gemini post does not say whether the regional voices run on the same text-to-speech models, so the comparison shows scale, not shared technology.
The same report noted that the Live API did not support the Flash TTS models, so real-time voice conversations would need a separate text-to-speech step. The Gemini app’s Live mode is a different surface from the API, which is why the post can describe regional voices in a live conversation. It does not explain how that is done.
For a reader who already uses Gemini by voice, the change is practical. Before this, the choice was about the character of a voice. Now it is also about where that voice sounds like it comes from. For people who use Gemini Live to practice a language, an accent that matches the target region may matter more than it does in casual chat. Google makes no claim about learning outcomes, and the post offers no test data.
What Google has not said
Everything in the announcement is Google’s own description. The claims that the voices sound natural, avoid exaggeration and capture local pitch and inflection have not been independently checked, and the post includes no audio samples in the text. There is no information on how the voices were produced, whether human voice actors were involved, or how the synthetic speech is labeled.
Other open points include supported countries, whether Workspace and enterprise accounts get the voices, and whether more languages will follow. The post does not say. Readers who want to judge the result can open the voice menu in the app and listen, which is the only test the announcement itself suggests.
Sources
Sign Up for the Daily AI Pulse
One email a day. All the stories that matter.
By signing up, you agree to our Terms of Use and acknowledge the data practices in our Privacy Policy. You may unsubscribe at any time.
Alex CarterAI News & Big Tech CorrespondentAlex Carter writes for Aiholics, keeping readers updated on the fast-paced world of AI and Big Tech.



