Sections

Search

ArXiv paper measures how much benchmark contamination actually inflates scoresGoogle's Gemini 3.8 TTS models are cheap and handle multi-voice dialogue, per Simon WillisonUniDataAgent: China Unicom's Ontology-Grounded Enterprise Q&A Agent Cuts Report Time from Days to MinutesOpenAI says Harvey uses GPT-6 Astra to produce more structured legal draftsArXiv paper proposes auditable LLM labeling for classroom talk
All stories

Models·

Google's Gemini 3.8 TTS models are cheap and handle multi-voice dialogue, per Simon Willison

Simon Willison reports Google's new Gemini 3.8 TTS models are very cheap and can generate conversations between multiple voices.

What happened

- Developer Simon Willison says Google's new Gemini 3.8 text-to-speech models are super-cheap and can generate conversations between multiple voices, drawing on a library of 2,000-plus voices, with the option to clone your own (Simon Willison on Bluesky). That voice count and cloning option are his description of the models' capabilities, not an independent test. He built a small playground UI for the models and used Claude to write a script in which two pelicans debate moving to Pacifica Pier (Simon Willison on Bluesky). [1]

Why it matters

- Multi-voice generation at low cost is the kind of capability that makes AI audio production and conversational voice apps cheaper to build; the reactions here are one developer's hands-on impressions rather than benchmark results. [1]

Sources

  1. Simon Willison (Bluesky) · Reporting ·
    The new Gemini 3.8 TTS models are super-cheap and can generate conversations between multiple voices (from 2,000+, or...