Google expands global language AI to support over 300 languages

Google announced new advancements in global language AI, supporting over 300 languages and reaching 7.1 billion people. The company emphasized the need for AI to understand human communication beyond simple translation, incorporating tone, emotion, and cultural context. Google introduced Gemini 3.5 Live Translate and Transcribe, offering real-time spoken translation across 70 languages and high-precision speech-to-text capabilities. To address low-resource languages, Google launched the '1,000 Languages Initiative,' leveraging the Universal Speech Model (USM) trained on 12 million hours of audio. Collaborations with local experts and institutions produced datasets like WAXAL, Project Vaani, and the Amplify Initiative, covering 27 languages in sub-Saharan Africa and 109 languages globally. Google also unveiled Language Explorer, a tool to visualize and map over 7,000 languages. Google developed lightweight models like TranslateGemma for edge devices and piloted the 'Ask Viamo Anything' (AVA) assistant in Rwanda, answering over 2 million questions. The company also introduced Sign Language to Text (SL2T) technology, supporting over 50 sign languages, and enhanced Māori language accuracy in maps. These efforts aim to expand global expression while respecting cultural contexts.