Google released Gemini 3.8 Flash TTS and Flash-Lite TTS on September 23, giving creators more control over how generated speech is performed. Scripts can specify delivery for individual lines, including pacing, emotion and conversational reactions.
Flash TTS supports designing an original voice from a written description and saving it for later projects. Both models can stage a two-speaker conversation from one script. Flash-Lite is aimed at higher-volume work such as dubbing and voice-agent output, while Flash emphasizes voice design and character performance.
Google also offers voice replication from a 30-second reference sample. Users must supply a matching consent recording from the speaker, and the reference must be their own voice or one they have rights to use. Generated audio carries Google’s SynthID watermark.
The developer rollout begins in Google AI Studio and the Gemini API. The announcement lists enterprise API access as coming later, and voice remixing as a future feature. Creators can use the Studio audio playground to design voices and direct dialogue.