Google DeepMind has launched Gemini 3.1 Flash TTS, its most controllable text-to-speech model yet.

New text-based audio tags let users direct vocal style, pace, pauses and delivery.