Google announces Gemini 3.5 Live Translate for instant voice-to-voice translation (arstechnica.com)

Google has unveiled Gemini 3.5 Live Translate, a cutting-edge model designed for instant voice-to-voice translation across more than 70 languages. This significant advancement marks a key evolution in Google’s ongoing pursuit of real-time translation, moving beyond previous limitations that required specific hardware setups and focusing on lower latency and wider accessibility. The model can keep pace with natural conversation, providing translations that closely mimic the original speaker's intonation, pacing, and pitch, ensuring a more natural interaction rather than a robotic output. The Gemini 3.5 Live Translate model is now integrated into various facets of the Google ecosystem, and developers can utilize it through the Gemini Live API or AI Studio in a public preview. With its ability to continuously process speech and automatically manage multilingual inputs, the model greatly reduces the manual configuration typically required for translation tools. Additionally, its advanced noise-cancellation features make it particularly effective in crowded environments, potentially transforming how individuals and businesses engage across language barriers in real time.