AI-generated editorial illustration; not a documentary photograph.
Google has detailed a broad expansion of its language AI, including Gemini 3.5 tools for live translation and transcription, lightweight TranslateGemma models, new speech datasets and sign-language support. The company says its products now handle interactions in more than 300 languages used by over 7 billion people, while Google Translate supports more than 250 languages.
Speech Beyond Transcripts
Gemini 3.5 Live Translate provides real-time spoken translation across 70 languages and more than 2,000 language pairs. Google says it can retain emotional cues and handle code-switching, where speakers move between languages during a conversation. Gemini 3.5 Transcribe converts audio into formatted text and is designed for noisy settings and specialized terminology.
The transcription model also powers Rambler in Android Gboard. According to Google, the feature removes filler words, corrects grammar and punctuation, accepts voice-based editing commands and supports switching between languages.
Data for Underrepresented Languages
Google is pursuing support for the world's 1,000 most-spoken languages through systems including its Universal Speech Model, which was trained on 12 million hours of audio. The company says cross-lingual transfer lets the model apply patterns from data-rich languages to languages with less training material.
Its associated data projects include WAXAL, an open speech dataset spanning 27 Sub-Saharan African languages, and Project Vaani, which has collected more than 30,000 hours of speech from over 155,000 speakers across 109 Indian languages. A new interactive tool, Language Explorer, visualizes the open-source LinguaMeta repository covering more than 7,000 spoken, written and signed languages.
Offline and Feature-Phone Access
TranslateGemma is a family of lightweight open translation models derived from Gemini and trained across 55 languages. Google says the models can run on-device, allowing translation without cloud or internet access, although capable hardware is still required.
For standard feature phones, Google is supporting Viamo's Ask Viamo Anything voice assistant. Viamo piloted the service in Rwanda through its interactive voice response system, and Google says Gemini has answered more than 2 million questions through it.
Sign-Language Support
Google's Sign Language-to-Text system was trained across more than 50 sign languages. It powers sign-to-text dictation in Gboard and Live Transcribe on Pixel 11, initially supporting American Sign Language to English. Google presents this as a first step rather than comprehensive support for every sign language included in training.
Source: AI for everyone in every language
SIGNAL TITANS
Leave a Reply