Google has released Gemini 3.8 Flash TTS and Flash-Lite TTS, new text-to-speech models that can design custom voices from natural-language prompts.

The launch was reported on September 24. The models let developers describe the voice they want, such as “a warm, calm narrator,” and generate it on the spot. They support more than 100 languages and come with a library of over 2,000 production-ready voices.

The models can clone a voice from just 30 seconds of audio, but only with the speaker’s verbal consent, Google said. All generated audio carries a SynthID watermark so it can be detected as AI-made. The models also attach C2PA credentials, a tamper-resistant record of how the audio was created.

Flash TTS is aimed at high-quality production use, while Flash-Lite TTS is a smaller, faster model for real-time applications like voice assistants and live translation.

The release comes as voice AI becomes a major battleground. Google is competing with OpenAI, ElevenLabs and others to power the next generation of AI assistants, audiobooks and dubbing tools.