Gemini 3.8 text-to-speech models released with 2,000 voices

Google released two new text-to-speech models today, gemini-3.8-flash-tts and gemini-3.8-flash-lite-tts, according to a report by Simon Willison. The announcement, posted to Google’s blog under its Gemini models and research section, describes a library exceeding 2,000 voices available to developers.

Willison notes the models include the ability to create a custom voice, though the excerpt available does not specify the full mechanism by which custom voices are generated or trained.

Willison also published a companion tool, the Gemini 3.8 TTS Playground, hosted on his own tools.simonwillison.net domain, allowing users to experiment with the new voice library directly.

The report identifies two distinct model tiers, a standard “flash” version and a “flash-lite” version, consistent with Google’s established pattern of offering both fuller-featured and lighter, presumably faster or cheaper, variants of its models. No further technical specifications, pricing details, or performance benchmarks are included in the available material.

The scale of the voice library, over 2,000 options, marks a substantial expansion in the variety of synthetic speech available through a single model family. The source material does not elaborate on licensing terms, availability by region, or the timeline for broader developer access beyond the initial release and the accompanying playground tool.