Google Introduces Gemini 3.8 Live with Interactive Avatars and New Flash TTS Models
Something you can actually use or run today.
On September 24, 2026, Google launched Gemini 3.8 Live, introducing real-time visual avatars, a low-latency TTS engine, and an Extended Thinking mode for complex reasoning.
It pushes the interface for conversational AI beyond voice into interactive video, though the high compute costs currently gate this to enterprise users.
Interactive avatars are the '3D TV' of AI—technically impressive but currently lacking a killer application that justifies the massive compute overhead. The 'Extended Thinking' mode is the real value here, mimicking OpenAI's reasoning focus.
Release of a 'Flash' version of the Live Avatar API for consumer-grade applications.
Also covers
- Google Releases Gemini 3.8 Flash and Flash-Lite Tts ModelsOn September 23, 2026, Google officially announced the release of Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS models.
- Shifts conversational AI from purely voice-only modes to interactive visual avatars.
- Currently limited to Enterprise environments, hinting at high compute requirements for simultaneous video-audio orchestration.
Sources agree that the feature provides a highly interactive frontend layer for conversational AI but note limited initial availability.