Simon Willison vibe-codes a Gemini 3.8 TTS playground: 1m18s audio for 2.74¢

Simon Willison · rss · 2026-09-24

Google released two new text-to-speech models, gemini-3.8-flash-tts and gemini-3.8-flash-lite-tts, with 2,000+ preset voices and custom voice cloning from a 30-second sample. Simon Willison vibe-coded a bring-your-own-key playground with GPT-6 Astra, leveraging the Gemini API's open CORS policy. A standout API feature is easy multi-character conversations with distinct voices and style instructions — demoed with a two-pelican debate scripted by Claude 4.5 Opus. Generating 1m18s of audio took 20 seconds and cost 2.74 cents with Flash TTS.

Original post →

More from coding & agent

coding & agent channel →