Gemini 3.8 Flash TTS and Flash-Lite TTS ship with prompt-designed voices and 30-second voice cloning
Google Blog·medium signal
Google released two TTS models on 2026-09-23. They take line-by-line direction of emotion, pacing and dialect, stage two-speaker scenes natively, cover 100+ languages, and clone a voice from a 30-second sample after consent verification. Both are live in AI Studio, and Gemini API access is rolling out now. Google reports a score of 71.4 and first place on Hume AI's Voice Design Benchmark. Pricing is not disclosed yet.