Google releases Gemini 3.8 Flash TTS and Flash-Lite TTS: voice cloning from 30 seconds, with verified consent
In one sentence On September 23 Google made two text-to-speech models GA in the Gemini API: Flash TTS, with voice design from a text description and cloning from a 30-second sample, and Flash-Lite TTS for high volumes. Every clip carries a SynthID watermark; cloned voices also get C2PA credentials.
Google updated its speech synthesis with two new models. Gemini 3.8 Flash TTS is built for quality: audiobooks, podcasts, games, two-voice dialogue, with line-by-line acting directions and sounds such as laughs or sighs. Flash-Lite TTS is the cheaper version for high-volume dubbing and voice assistants.
Two features stand out. The first is designing a voice by describing it in words (role, accent, character), alongside a library of more than 2,000 ready-made voices. The second is cloning: a 30-second sample is enough, but it also requires a recording in which the voice owner gives consent, and the system checks it is the same person.
All generated audio carries an invisible SynthID watermark, and cloned voices also carry C2PA provenance credentials. The models are available in the Gemini API and Google AI Studio; enterprise access through Gemini Enterprise is announced as coming soon.
Companies
Tools
Gemini 3.8 Flash TTS, Gemini 3.8 Flash-Lite TTS, Google AI Studio
Tags
Sources