September 23, 2026
Voice for your app now comes from a prompt: Gemini speaks 100+ languages
On September 23, Google introduced two models, Gemini 3.8 Flash TTS and Flash-Lite TTS: they generate speech from text prompts in 100+ languages. The models are announced for Google AI Studio, Gemini Notebook, and Gemini API. They can conduct dialogue between two speakers, add laughter, sighs, and brief listener responses.
On September 17, the Gemini API documentation for speech generation still listed Gemini 3.1 Flash TTS Preview with a 32k-token session limit. Google now adds two 3.8 versions and claims first place in Hume AI Voice Design: 71.4 points, including 60.8 for accent modeling.
A voice for the task. A 30-second audio sample is enough for a permanent voice profile, provided the user holds rights to the voice. Google also announced a catalog of 2,000+ ready-made voices, and every clip receives a SynthID watermark.
The next milestone will be independent measurements of Gemini 3.8 Flash TTS quality and latency in the API.
