Gemini 3.8 text-to-speech says hello

Google has released an update to its Gemini text-to-speech model, version 3.8, which improves the model's ability to generate high-quality voices for a wide range of languages. The update brings the model's total vocabulary to 1.5 million words and 1.5 million phrases, allowing for more nuanced and context-specific speech. Gemini 3.8 also supports 26 languages, including many less common languages such as Welsh, Scottish Gaelic, and Irish. The model is trained on a dataset that includes over 500,000 hours of audio from publicly available sources, such as podcasts, audiobooks, and online lectures.