Technology

Google Adds 66 Regional Dialects to Gemini Voice on Android and iOS

Martin HollowayPublished 15m ago3 min readBased on 1 source
Reading level
Google Adds 66 Regional Dialects to Gemini Voice on Android and iOS
Photo by Czapp Árpád on Pexels

Google has added 66 regional dialects to Gemini, and the voices are available in the Android and iOS apps from 16 November 2026. Users can pick a dialect through voice settings in the Gemini app, or through the three-dot menu, according to Engadget.

The UK set is the most granular of the group. Gemini users in Britain can choose Geordie for Tyneside, Scouse for Merseyside and Brummie for Birmingham. The UK options also include Scots, Welsh, Northern Irish and Multicultural London English. Google has added tailored dialects for India, Brazil, Indonesia and Korea too, so the feature reaches well beyond the British Isles.

Google first previewed the feature at Google I/O earlier in 2026. The release moves it from announcement to availability on the two dominant mobile platforms, through the Gemini apps that many users already rely on for voice interaction.

According to the same report, Google worked with native speakers to make the dialect voices authentic and clear rather than exaggerated and stereotypical. That matters more than it first appears. Accent rendering in speech synthesis has a long record of caricature, where a few phonetic markers are overemphasised until the output sounds like parody. Getting a regional voice right means handling prosody (the rhythm, stress and intonation of speech), vowel shifts and timing, not just swapping a pronunciation table. Native-speaker review is the standard way to catch those failures, but it is labour-intensive and has to be repeated for every variant.

In my view, the scale of this release raises a practical question the announcement leaves open. Sixty-six dialect voices is a large maintenance surface. Each variant needs its own evaluation, regression testing as the underlying model changes, and its own quality bar. The reporting does not say whether these are fine-tuned voice models, adapters on a shared speech stack, or prompt-level steering of a single model. Developers building on Gemini's voice interfaces will want that detail before assuming the dialects behave consistently across updates.

The broader context here is user experience. Voice assistants have largely defaulted to a small number of standard accents, so speakers of regional English, or of the non-English languages covered by the India, Brazil, Indonesia and Korea options, have often heard a voice that sounds like it came from somewhere else. A voice that matches the listener's speech patterns lowers that friction. It also makes spoken interaction feel less like a distant service and more like talking to something nearby.

The obvious risk is the one Google's native-speaker process is meant to address. A regional accent rendered badly can feel more alienating than a neutral one, and a dialect voice that drifts into stereotype can cause offence. The release lists the dialects and markets but does not say how errors will be reported or fixed, and that will shape whether the feature earns trust over time.

In my view, the direction is still worth noting. For most of the history of consumer speech interfaces, choosing a voice meant picking the least objectionable default. Making local speech a configurable setting, across 66 variants and several language markets at once, treats regional identity as something software should accommodate. If the quality holds up under daily use, that is a meaningful step toward voice interfaces that sound like the people who use them.