Gemini 3.8 TTS Playground
Simon Willison’s Weblog Simon Willison ● Covered by 2 sources
Google shipped two new Gemini text-to-speech models, plus a playground for custom voices. The odd bit: you can make a voice from a 30-second sample, and 1m 18s of audio cost 2.74 cents.
Based on reporting by Simon Willison’s Weblog, Simon Willison — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Google released two new text-to-speech models under the Gemini name: gemini-3.8-flash-tts and gemini-3.8-flash-lite-tts. Alongside them comes a voice library with more than 2,000 options, and Google says custom voices can be created from just a 30-second audio sample of your voice, or one you have the rights to use.
Simon Willison built a bring-your-own-key playground for the models, using GPT-6 Astra and the Gemini API’s open CORS policy. The point of the tool is not just to spit out a single speaker. The API also makes it easy to script conversations with multiple characters, each with its own voice and style instructions.
Willison used that setup to make a demo conversation between two pelicans arguing about whether they should move to the Pacifica Pier. Claude 4.5 Opus wrote the script and generated a URL for rendering it through the tool.
On the runtime side, the result was quick but not instant: about 20 seconds to generate 1 minute and 18 seconds of audio with Gemini 3.8 Flash TTS, the non-Lite model. The cost came out to 2.74 cents.
My take — AI-written commentary, not fact-checked reporting
This is the kind of AI feature that actually feels useful instead of theatrical. A huge voice library, short samples for custom voices, and multi-character dialogue make more sense than yet another chatbot demo pretending to be a product. The only surprise is that the practical stuff keeps being the least flashy part of the pitch.
Read more about this at: Simon Willison’s Weblog