Google DeepMind has released Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, two new audio models for creating personalized voices across more than 100 languages with a library of 2,000 ready-to-use options. The Flash model targets high-fidelity creative projects like gaming and podcasts with line-by-line control and natural audio cues, while the Flash-Lite model is built for efficiency in real-time voice agents. Upon release, the Gemini 3.8 Flash TTS model took the top spot on the Artificial Analysis Pronunciation Robustness benchmark with a score of 89.5%, surpassing the previous leader, Gemini 3.1 Flash TTS.

Pricing for the new tools is $32.98 per 1M characters for the Flash model and $22.07 per 1M characters for Flash-Lite, which is more expensive than Gemini 3.1's $18.31 but lower than Eleven v3 at $100. In terms of performance, Flash TTS processes 44.1 characters per second, or 2.7x faster than real-time, though it lags behind competitors like Falcon 2, which reaches 204.9 characters per second. The model also debuted at No. 2 in the Provider Voice Arena with an Elo of 1,263, trailing only Sonic 3.6.