Introducing Gemini 3.8 Live with Live AvatarEN
Aggiunta all'archivio di AIChronos il (numero del 24 settembre 2026 ricostruito)· fonte del · Google DeepMind
Google DeepMind introduces Gemini 3.8 Live with Live Avatar, an enterprise feature that natively couples low-latency streaming video with live dialogue models to provide near real-time visual presence. The system processes audio and visual inputs simultaneously to generate expressive responses featuring precise lip-syncing, natural facial expressions, and fluid turn-taking across 97 languages using native multilingual speech-to-speech synchronization. Powered by asynchronous tool calling, the avatar can execute background tasks while maintaining an uninterrupted conversational flow for complex workflows.
- Live Avatar enables dynamic visual personas that listen, see, and speak with precise lip-syncing to enhance enterprise customer service and walkthroughs.
- The feature supports native multilingual synchronization across 97 languages without degrading video fidelity or introducing visual drift.
- Asynchronous tool calling allows the model to fetch data and trigger actions in the background while maintaining active dialogue continuity.
- Organizations can customize avatars from high-quality reference images, though custom creation is currently restricted via enterprise allowlisting.
Why it matters. Developers building LLM agents or local AI systems gain a new paradigm for multimodal interaction where visual presence and asynchronous reasoning coexist without interrupting the user experience. This capability allows for the construction of more immersive digital exchanges that can handle complex, multi-step tasks while maintaining strict identity transparency through SynthID watermarking.
Scheda generata automaticamente da un modello AI locale a partire dalla fonte originale: può contenere imprecisioni.