Gemini 3.8 Flash TTS adds custom voice creation from recordings or a sentence

Gemini 3.8 Flash TTS and Flash-Lite TTS are now available in the Gemini API and AI Studio, with a new option to replicate a user's own voice from two recordings or design one from a sentence.

The guide says the reusable voice ID can be passed in later requests, or an encrypted voicekey that expires after 7 days can be used if nothing is stored server-side.

Prompting changed from gemini-3.1-flash-tts-preview: input text is spoken word for word, delivery goes in speech_metadata.style, and non-streaming responses are now real WAV.