Gemini 3.5 Live Translate (blog.google)

Google has unveiled Gemini 3.5 Live Translate, an advanced speech-to-speech translation model capable of seamlessly translating over 70 languages. This model introduces continuous translation, allowing it to interpret and generate speech while the speaker is talking, significantly enhancing the user experience by eliminating awkward pauses typical in previous systems. The new capabilities are geared for use across various Google products, including Google Meet and the Google Translate app on both Android and iOS, enabling more natural conversations in multilingual settings. This release is significant for the AI/ML community as it represents a major leap forward in real-time translation technology, echoing advancements in natural language processing. Gemini 3.5 brings noise robustness and is designed to handle multilingual inputs dynamically, meaning users no longer need to adjust settings for different languages. Additionally, the incorporation of a watermarking technique called SynthID ensures that AI-generated content is traceable, bolstering efforts to combat misinformation. With early positive feedback from various partners, including Grab, this model promises to facilitate smoother, more effective communication in real-time applications like meetings and calls, making it a transformative tool in breaking down language barriers.