Gemini 3.8 text-to-speech

Google has introduced two new text-to-speech models, dubbed Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, which bring a heightened level of customization and expressiveness to audio generation. These models empower creators, developers, and enterprises to craft unique, natural-sounding voices tailored to their specific needs, from character voices for audiobooks to brand ambassadors for podcasts.

Both TTS models offer robust control over voice delivery, securing top spots on leading quality benchmarks. Gemini 3.8 Flash TTS, for instance, outshines competitors with a Hume AI Voice Design Benchmark score of 71.4, while Gemini 3.8 Flash-Lite TTS ranks second in overall quality.

These advancements are complemented by the Gemini Audio family, which already includes models like 3.5 Live Translate, 3.5 Transcribe, 3.8 Live, and 3.8 Live Extended Thinking. With the ability to scale from 30 original voices to an infinite library, users can now create richly expressive audio experiences, whether they're building an entirely new character or fine-tuning a consistent brand voice.

To ensure responsible use, Google has implemented strict safeguards. For voice replication, users must provide a verbal consent recording from the voice owner, and every audio clip generated by Gemini Audio models is watermarked with SynthID to prevent misinformation.

These new models are now available in Google AI Studio, allowing developers to build and deploy speech generation experiences with ease. Companies like Figma, HeyGen, Linguana, Wondercraft, 99.co, and Ollang have already integrated these TTS models, enhancing global dubbing, localization, and conversational voice agents. However, it's worth noting that voice replication through AI Studio is currently unavailable in certain regions such as Illinois, Texas, EEA, UK, Switzerland, and India.