Gemini 3.8 Flash TTS opens voice design from a written prompt
Google opened Gemini 3.8 Flash TTS and Flash-Lite TTS in the Gemini API and Google AI Studio on 23 September. Flash TTS designs a voice from a written prompt and copies one from a 30-second sample after a spoken consent check. Enterprise API access has not started. A ready-made library sits above 2,000 voices.
Artificial Intelligence··Night
Two speech models open in the public tools
Google introduced Gemini 3.8 Flash TTS and Flash-Lite TTS on 23 September, and both are opening in the Gemini API and Google AI Studio. Flash TTS is the model for line-by-line performance direction. Flash-Lite TTS is the model for high-volume dubbing. Flash TTS is also opening in Gemini Notebook, and Flash-Lite TTS is opening in Google Vids. The same Wednesday release, as The Next Web reports it, puts both models in the Gemini API and Google AI Studio and treats Flash TTS as the creative tool for games, audiobooks and podcasts, with Flash-Lite TTS kept for the higher-volume work. The public developer path is what opened. A reader who only saw a product name would miss the split: one model takes a written direction for a performance, the other is built to dub at volume, and both sit in the same public tools on the day of the announcement.[1], [2]
A prompt designs the voice, and a sample can copy one
Flash TTS offers voice replication from a 30-second sample after a consent check. The sample has to be one the user has rights to, and a spoken consent recording has to accompany it. Google says it adds a SynthID watermark and C2PA credentials to that copy. Google states coverage of more than 100 languages. Replication through AI Studio is unavailable in Illinois, Texas, the European Economic Area, the United Kingdom, Switzerland and India. The Next Web describes the same 30-second route and says the consent recording is checked against the reference speaker before the voice is created. It also puts a ready-made library above 2,000 voices beside the designer. Voice design from a text prompt is the other route on both accounts: the user describes the voice, and the model builds it, rather than only picking from the library.[1], [2]
The enterprise door stays shut
The company says Gemini Enterprise API access has not started. Voice remixing is listed as not yet available, so timbre, pitch, pace and accent are not controls a developer can turn on Wednesday. The Next Web's account of the release does not open that enterprise door either: it reports the public models, the prompt, the 30-second copy and the library above 2,000 voices. What a developer can use on 23 September is therefore the public pair, Flash TTS for directed performance and Flash-Lite TTS for high-volume dubbing, inside the Gemini API, Google AI Studio, Gemini Notebook and Google Vids. The geographic block on AI Studio replication still stands in Illinois, Texas, the European Economic Area, the United Kingdom, Switzerland and India. The consented 30-second sample, the spoken consent recording, the SynthID watermark and the C2PA credentials are the conditions on the copy that did open.[1], [2]
