Google introduced Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS on September 23. The company positions Flash TTS for detailed creative direction and character design, while Flash-Lite targets higher-volume, lower-cost speech generation.
Custom voices and directed performances
Google says the models can create voices from natural-language descriptions, direct delivery line by line and maintain characters across long-form audio. It also describes consent verification for voice replication, SynthID watermarking and C2PA credentials as safeguards for generated speech.
Available through Google’s model platforms
Google says the models are available through Gemini API and Google AI Studio, with integrations across Gemini Enterprise, Gemini Notebook and Google Vids. The associated model card says the Gemini 3.8 Audio family is based on Gemini 3 Pro and notes familiar foundation-model risks including hallucinations, jailbreaks and occasional latency or timeouts.
Google reports top positions on Hume AI and Voice Arena evaluations, but those results were not independently reproduced for this article. Real-world voice quality, consent enforcement and cost efficiency will depend on deployment conditions and user behavior.