Google DeepMind Releases Gemini 3.8 Flash TTS and Flash-Lite TTS

Google DeepMind has introduced two text-to-speech models, Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, designed to scale up from static presets to an infinite library of custom voices. Released on Sep 23, 2026, the new models let creators direct accents, emotional tones, and line-by-line dialogue performances.

Gemini 3.8 Flash TTS secured the 1st overall spot on Hume AI's Voice Design Benchmark with a score of 71.4, while also leading in accent modeling with 60.8. Both models secured the 1st and 2nd spots respectively on Hume AI's Overall Quality Index. In blind human preference evaluations on Voice Arena, the models scored top positions across global languages including Japanese, Brazilian Portuguese, Vietnamese, Modern Standard Arabic, Mexican Spanish, and Hindi, with support extending to over 100 languages.

The audio generation models include safety mechanisms such as SynthID watermarking embedded directly into outputs. Voice replication requires consent verification through a verbal recording from the voice owner matching the reference speaker.

Developers can access the speech generation tools within Google AI Studio and build via the Gemini API, with integrations in Gemini Notebook and Google Vids. Voice replication through AI Studio excludes Illinois, Texas, the European Economic Area, the United Kingdom, Switzerland, and India.