AI for everyone in every language
Google says its technologies now support more than 300 languages, and it is shifting from text translation toward models that process speech directly, capturing tone, emotion and code-switching. It is also building open datasets with local communities, offering on-device and feature-phone tools for low-connectivity areas, and adding sign-language-to-text features.
What the article says
In a post dated September 15, 2026, James Manyika, a senior vice president at Google, says the company's technologies now support everyday interactions in more than 300 languages, spoken by over 7 billion people, or 86 percent of the world's population. The post notes that technology has mostly served a few dominant languages and that thousands of living languages remain poorly represented online. It describes a shift from text translation toward models that process speech directly, including tone and mixed-language speech.
Announcements include Gemini 3.5 Live Translate (real-time spoken translation across 70 languages and more than 2,000 pairs), Gemini 3.5 Transcribe for noisy-environment transcription, TranslateGemma (open translation models that run offline, trained across 55 languages), a sign-language-to-text tool, initially American Sign Language to English, and Language Explorer, which maps more than 7,000 languages. Google also cites the WAXAL dataset (27 Sub-Saharan African languages) and Project Vaani (over 30,000 hours of speech in 109 languages). The post gives few availability dates, and the 1,000-language goal remains an aim rather than a current capability.
Read the official article AnnuIA profile
Summary written by AnnuIA in its own words from the official page of Google (IA). The content and rights belong to the publisher: please read the original article.