Google dropped two new Gemini 3.8 Flash text-to-speech models with 2,000+ stock voices across 100+ languages. They can now generate natural audio cues like laughs, sighs, and "mhm".
Game-changer for production and development, or a fast route to uncanny valley?
