Google ships Gemini 3.8 Flash TTS and Flash-Lite TTS, then adds Live Avatar faces to Gemini 3.8 Live

Google DeepMind released Gemini 3.8 Flash TTS and Flash-Lite TTS with prompt-based voice design, a 2,000-plus voice library, and consent-gated cloning, then on 24 September 2026 added Live Avatar, lip-synced real-time video avatars for Gemini Enterprise.

On 23 September 2026, Google DeepMind released Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS (model IDs gemini-3.8-flash-tts and gemini-3.8-flash-lite-tts). Flash TTS generates voices from text prompts describing role, accent, and traits, supports line-level performance control and stage directions, and via the Gemini API can stage scripted conversations among multiple characters, each with its own voice and style instructions; Flash-Lite targets high-volume, cost-efficient production such as dubbing and voice agents. Both cover more than 100 languages (MarkTechPost adds dialects) and a library of more than 2,000 production voices, which MarkTechPost says is up from 30. Voice replication works from a 30-second sample the user has rights to; consent is described as verification (DeepMind, Hacker News), a spoken consent statement (The Decoder), or a verbal consent recording matched to the owner (MarkTechPost). All audio carries SynthID watermarks, with C2PA credentials reported for all output by DeepMind and Hacker News and specifically for replicated voices by MarkTechPost. Flash TTS ranks first on Hume AI's Voice Design Benchmark at 71.4 and first on the Overall Quality Index (DeepMind, Hacker News), with Flash-Lite second (Hacker News, MarkTechPost); The Decoder says Google claims leads on most Hume TTS categories. Access accounts differ: DeepMind and Hacker News list Google AI Studio, the Gemini API, Gemini Enterprise, Gemini Notebook, and Google Vids, while The Decoder and MarkTechPost describe a Gemini API and AI Studio rollout, with MarkTechPost calling it API-only with no open weights and Gemini Enterprise access coming soon. The Decoder prices one audio hour at about $0.81 for Flash TTS and $0.54 for Flash-Lite through the end of 2026, rising to $1.62 and $1.08 in 2027. Separately, Simon Willison published a bring-your-own-key playground built with GPT-6 Astra that uses the API's open CORS policy. On 24 September 2026, Google DeepMind announced Gemini 3.8 Live with Live Avatar, available in Gemini Enterprise. Live Avatar pairs near real-time video with speech for lip-synced avatars whose expressions adapt across 97 languages without claimed loss of video fidelity, can show information on screen while speaking, and supports asynchronous tool calls that fetch data in the background without interrupting dialogue. Enterprises can pick preset avatars, and organizations on an allowlist can generate custom avatars from a reference image. Audio and…

• Models released 2026-09-23: Gemini 3.8 Flash TTS (gemini-3.8-flash-tts) and Gemini 3.8 Flash-Lite TTS (gemini-3.8-flash-lite-tts).
• Voice library exceeds 2,000 production voices, up from 30 per MarkTechPost.
• Voice replication requires a 30-second sample the user has rights to; consent is variously described as verification, a spoken consent statement, or a matching verbal consent recording.
• Both models support 100-plus languages (MarkTechPost adds dialects) and line-level stage directions; the Gemini API supports multi-character conversations with per-voice style instructions.
• Flash TTS ranks #1 on Hume AI's Voice Design Benchmark at 71.4 and #1 on the Overall Quality Index; Flash-Lite ranks #2 in overall quality; The Decoder says Google claims leads on most Hume TTS categories.
• Safeguards: SynthID watermarking on all audio; C2PA credentials reported for all output by DeepMind and Hacker News but only for replicated voices by MarkTechPost.
• Pricing (The Decoder): roughly $0.81 per audio hour on Flash TTS and $0.54 on Flash-Lite through end of 2026, rising to $1.62 and $1.08 in 2027.
• Access lists conflict: DeepMind and Hacker News name Google AI Studio, the Gemini API, Gemini Enterprise, Gemini Notebook, and Google Vids; The Decoder and MarkTechPost describe API and AI Studio rollout, with MarkTechPost calling it…