Gemini 3.8 Launch Details

On September 23, 2026, Google introduced the Gemini 3.8 Flash TTS and Flash-Lite TTS models, featuring over 2,000 production-ready voices and support for more than 100 languages. The models achieved a Voice Design Benchmark score of 71.4 and an Accent Modeling score of 60.8, ranking #1 and #2 in the Overall Quality Index for Flash TTS and Flash-Lite TTS, respectively.

The Gemini 3.8 models are designed for scalability, with an infinite library of original voices, and can replicate audio samples of up to 30 seconds in duration. Google emphasizes that each audio clip generated by these models is watermarked with SynthID, ensuring traceability.

Leland Rechis, Group Product Manager, and Alan Cowen, Director of Research Science, highlighted that the voice creation and replication capabilities were built with strict safeguards, reflecting Google's commitment to responsible AI development.

The launch follows the earlier release of Gemini 3.5 Live Translate and Transcribe models, further expanding Google's audio generation offerings. Partner companies such as Figma, HeyGen, and Linguana are expected to integrate these new capabilities into their platforms, potentially enhancing user experiences in audio content creation.

Developers can access these features through the Gemini API, which is part of the broader Google AI Studio platform for audio generation.