Google launches Flash TTS voice models

Google introduced Gemini 3.8 Flash TTS and Flash-Lite TTS for speech generation. It said Flash TTS was for creative work such as podcasts, audiobooks and game characters, while Flash-Lite TTS was for lower-cost speech generation at scale for dubbing, audio content and voice agents.

For Flash TTS, Google said users could describe a voice’s role, accent and other traits in text. It also said the service included more than 2,000 preset voices, including regional variants such as Mexican Spanish, Quebec French and Scottish English.

Google said voice cloning used a 30-second audio sample and required a spoken statement of consent from the person being cloned. It said generated clips carried an inaudible SynthID watermark, and that Voice Remixing would later let users alter timbre, pitch, tempo and accent.

Both models could take line-by-line directions or interpret script cues, and Google said they could generate two-voice dialogue plus sounds such as laughter and sighs. In the article’s tests, a style prompt produced a German-accented voice, but both clips included a high-pitched background whine and one ended with a change in the voice.

Google said the models were rolling out through the Gemini API and Google AI Studio, with Gemini Notebook and Google Vids also carrying access. It said Gemini Enterprise API access would follow, and pricing was listed per million tokens, with one hour of audio equal to 90,000 audio tokens.