What's this about?
Gemini Enterprise can now stream a visual persona that sees, listens and speaks while maintaining lip-sync and facial expression. Tool calls can execute asynchronously without freezing the conversation.
What exactly is native multimodal streaming?
Audio, video and conversational context are processed as one continuous exchange instead of gluing a chatbot, speech engine and pre-rendered face together after the fact.
Why does this matter?
official launch and model card; independent usage evidence was not yet available in the collection window.
- The system can switch among 97 languages while adapting mouth movement.
- Custom likenesses require enterprise allowlisting; presets are available more broadly.
- SynthID marks output, but live human-like avatars still raise impersonation and disclosure questions.