/tech

Google launches Gemini 3.8 Flash TTS and Flash-Lite TTS

Google says Flash TTS can create custom voices from text prompts across more than 100 languages and dialects, while Flash-Lite targets high-volume dubbing, audio creation and voice agents.

TagsGoogle Google Logan Kilpatrick

🚨A

4 Sources, ,

Google launch graphic introducing Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS on a pale blue background.

TLDR

Google announced the two text-to-speech models on September 23, 2026, with rollout beginning that day in the Gemini API and Google AI Studio. The company lists a library of 2,000-plus voices and says both models support line-by-line performance direction and two-speaker scenes. Voice remixing was listed as coming soon.

Google says voice replication uses a 30-second sample of your own voice or one you have rights to use, and requires a matching verbal consent recording from the voice owner. It says generated audio carries SynthID watermarks. Unite.AI reports that voice replication through AI Studio is unavailable in Illinois, Texas, the European Economic Area, the UK, Switzerland and India.

Featured Source

Google has released two Gemini text-to-speech models, Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, for developers using the Gemini API and Google AI Studio. In its announcement, the company describes Flash TTS as the higher-capability model and Flash-Lite as the lower-cost option for high-volume work such as dubbing and voice agents.

Both models are built around instructions for delivery, not just the words to be spoken. Google says developers can direct a line’s pace, tone and style, and can create two-speaker scenes. The company also says the release supports more than 100 languages and dialects, a voice-design workflow based on text prompts, and a catalog of more than 2,000 production-ready voices.

What the two models are for

The distinction is primarily about the trade-off between expressive output and scale. Flash TTS is intended for more nuanced delivery; Flash-Lite TTS is positioned for applications that need to generate a large amount of audio efficiently. Those are product claims from Google, rather than an independent comparison of quality or cost.

Google is also introducing voice replication. Its announcement says that feature uses a 30-second sample and requires a matching verbal-consent recording from the voice owner. The company says users must have the right to use the voice and that generated audio carries SynthID watermarking.

Availability still has boundaries

The models began rolling out on Sept. 23 through the Gemini API and Google AI Studio. Google lists voice remixing as a future feature, not part of the initial release. Availability can also vary by product and location: Unite.AI reported that voice replication in AI Studio is unavailable in Illinois, Texas, the European Economic Area, the UK, Switzerland and India.

For developers, the immediate change is a new pair of API-accessible speech models with more controls over how a line is performed. The broader questions, including how well those controls hold up across languages and real production workloads, will require testing beyond Google’s launch materials.

TLDR

What the two models are for

Availability still has boundaries

Google launches Gemini 3.8 Flash TTS and Flash-Lite TTS

TLDR

What the two models are for

Availability still has boundaries

Google launches Gemini 3.8 Flash TTS and Flash-Lite TTS

6 Sources

Combined views

What the two models are for

Availability still has boundaries

Combined views

Sentiment

Summary

TLDR

What the two models are for

Availability still has boundaries

Sentiment

Summary

Sentiment

Summary

What the two models are for

Availability still has boundaries

6 Sources

Sentiment

Summary