Celebrating 20 years of Google Translate: Fun facts, tips and new features to try
models
read at source ↗ blog.google
Celebrating 20 years of Google Translate: Fun facts, tips and new features to try
Source: Google Date: 2026-04-28 URL: https://blog.google/products-and-platforms/products/translate/fun-facts-google-translate-20-years/
Summary
Google Translate turns 20 with 1 billion monthly users translating roughly 1 trillion words per month across nearly 250 languages. The milestone release adds AI-powered pronunciation practice on Android (English, Spanish, Hindi) — described as the most-requested feature — powered by Gemini models. The service has evolved from 2006 statistical machine translation to neural MT in 2016 to current Gemini integration.
Implications
- Scale benchmark for language AI. 1 trillion words/month is a useful reference point for what consumer-grade language AI infrastructure looks like at saturation. The shift to Gemini models means Google is now routing production translation workloads through its frontier model stack rather than a specialized MT pipeline.
- Pronunciation feedback as a wedge into language learning. Duolingo’s moat has been interactive speaking practice; Google adding free AI pronunciation feedback with 1B users as the distribution surface is a direct pressure on the paid language-learning market.
- Feeds the multimodal / ambient interface thread: Live real-time audio-to-audio translation (already shipped in Lens/earbuds) is the clearest consumer demonstration of the ambient translation layer — voice as interface, language barrier removed. The 5-minute+ session duration metric suggests users are leaning into sustained conversations, not just phrase lookups.