How does Google make AI understand 300 languages and still miss billions of people?

The milestone was announced on September 15, 2026. The post comes from James Manyika, SVP of Research, Labs, Technology & Society.

From text pipelines to native audio intelligence

For decades speech recognition followed a rigid chain: transcribe audio to text, process the text, synthesize it back to audio. That pipeline strips away tone, pacing, emotion, and context. People do not speak in perfectly neat grammatical sentences. Real speech is messy. They laugh, overlap, hesitate, and weave multiple languages together mid-sentence. Spanglish and Hinglish are everyday realities.

The shift moves beyond text transcripts to native audio intelligence. Models like Gemini now process audio directly as is, while also grasping both sound and intent. The shift shows up in three concrete tools.

Gemini 3.5 Live Translate powers real-time spoken translation across 70 languages and 2,000+ language pairs. It naturally captures code-switching and emotional cues along the way. Gemini 3.5 Transcribe is the most precise speech-to-text model yet, turning raw audio into polished formatted text even in noisy environments or with complex jargon. It also powers Rambler on Android Gboard, which removes filler words, fixes grammar and punctuation, and lets you edit or rewrite with voice commands while switching seamlessly between languages.

The Universal Speech Model was trained on 12 million hours of audio. It uses cross-lingual transfer learning, techniques that enable models to transfer what they learn from data-rich languages to improve speech understanding in languages with far less training data. This allows models to apply patterns learned from data-rich languages to under-resourced ones. The work builds on 25 years of open research and more than 400 peer-reviewed speech papers.

Grassroots data partnerships replace web scraping

Because the web disproportionately represents a few dominant languages, teaching AI to understand underrepresented languages required rethinking data gathering. The solution is local grassroots partnerships. Three open-data partnerships illustrate the approach.

WAXAL — Wolof for "speaking," pronounced "Wah-hal" — is built with partners including Makerere University and Digital Umuganda. It is a large-scale open speech dataset covering 27 Sub-Saharan African languages spoken by more than 100 million people across more than 26 countries. The dataset captures tonal variation and conversational rhythms often missing from traditional datasets.

Project Vaani, in partnership with the Indian Institute of Science and Bhashini, maps India's linguistic diversity through a region-anchored rather than language-anchored approach. It has collected to date more than 30,000 hours of speech across 109 languages from more than 155,000 speakers.

The Amplify Initiative teamed up with more than 1,600 local experts and 20 universities across four continents, including Brazil's UFMG, India's IIT Kharagpur, and Uganda's Makerere University, to contribute 15,000 multimodal data points capturing local nuance.

An interactive tool called Language Explorer visualizes LinguaMeta, the world's largest open-source language data repository. Recognized by Fast Company for design innovation, it continuously maps more than 7,000 spoken, written, and signed languages.

Efforts supported by Google.org, including the Centre for Digital Language Inclusion and AI Singapore's Project Aquarium, are helping bring multilingual tools to farmers, healthcare workers, teachers, and other essential community members around the world.

Offline models and feature phones bridge the connectivity gap

For more than 3 billion people, reliable internet access is still out of reach. Technology is only truly accessible if it works where people live, including areas with limited or intermittent connectivity.

TranslateGemma is a family of lightweight open translation models built from Gemini and trained across 55 languages. Because TranslateGemma runs efficiently on-device, high quality translation no longer requires a connection to the cloud or the internet.

Still, running powerful AI models requires capable hardware, which excludes the hundreds of millions of people still using feature phones in low-resource regions. To bridge this divide, Google supports organizations like Viamo to power "Ask Viamo Anything" (AVA), a voice AI assistant that brings the power of Gemini to standard feature phones. Viamo has successfully piloted AVA in Rwanda with its existing interactive voice response users, and the service has already used Gemini to answer more than 2 million questions.

Accessibility and cultural pronunciation

Communication is not just about regional dialects or vocabulary. It is also about the many other ways people communicate. Conventional speech tools frequently fail people with non-standard speech, making them adapt to the technology rather than the other way around.

Sign Language-to-Text (SL2T) is trained across 50+ sign languages. It powers sign-to-text dictation in Gboard and Live Transcribe on Pixel 11, starting with American Sign Language (ASL) to English. This is an important first step toward making products more accessible to the 70 million people worldwide who rely on sign language to communicate.

Details matter in everyday tools like navigation. When a navigation app mispronounces a town or street name, it does not just cause confusion. It can overlook the cultural heritage of the place. In New Zealand, Google worked with Māori language experts to improve place name pronunciation in Google Maps, helping make the experience more accurate and genuinely local. Incorporating culturally authentic pronunciations directly into text-to-speech models ensures technology reflects the language and heritage of the communities it serves.

Where the tools live now

Today Google's language technologies are already embedded across its core ecosystem, connecting more than five billion people across nine platforms, including Search, Android, Chrome, YouTube, and Google Play. The goal is to support the world's 1,000 most-spoken languages. The company says it will continue to work closely with local communities to build technology that helps more people communicate, participate, and be understood on their own terms.

The announcement does not say how many of the 300-plus languages have full feature parity across all nine platforms. It does not specify which of the 1,000 target languages are next in line for native audio models. It does not give a timeline for expanding SL2T beyond ASL to English on Pixel 11.