Google has released two new text-to-speech models, Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, both shipping as stable rather than preview. Between them they cover two ends of the voice-synthesis market: one for quality, one for scale.
Flash TTS is the flagship. Google positions it for studio-grade fidelity, natural delivery and authentic regional accents, and says it holds up over long, multi-turn conversations. It supports 130 languages. Flash-Lite TTS trades some of that polish for speed and cost, aimed at high-volume production work and voice-agent pipelines; it covers 101 languages and replaces the older gemini-3.1-flash-tts-preview model.
The feature tying both together is prompt-based voice design. Instead of picking from a fixed list of voices, developers describe the voice they want in natural language, and Google AI Studio lets them design and refine that voice interactively before using it in an application. Both models share the same API schema and prompt structure, so switching between the flagship and the lighter model is a case of changing one parameter rather than rebuilding a pipeline.
The release fits a broader pattern among the large AI labs: speech is becoming a first-class, controllable output alongside text and images, rather than a bolt-on feature. For developers building voice agents, customer service bots or audio content tools, having a cheaper, faster sibling model with matching behaviour to the flagship removes a common reason to stay on an older, cruder TTS system.
Source: MarkTechPost
