Google Gemini introduces Live Avatar for conversational AI

The new feature is now generally available through Gemini Enterprise after being previewed at Google Cloud Next 2026. It enables conversational video agents to generate video avatars with synchronized lip movements across web, mobile and interactive kiosk applications.

Gemini 3.8 Live provides native speech-to-speech capabilities designed to support fluid dialogue, including more natural recovery from interruptions while maintaining conversation context and ongoing backend transactions.

The model can also execute tools and API calls in the background while continuing to interact with users, allowing it to acknowledge requests and maintain the conversation while tasks are being completed.

Gemini 3.8 Live can understand and speak 97 languages and automatically detect the language being used. Its live visual understanding capability allows the system to process live camera feeds and screen sharing alongside audio.

For identity protection and to help prevent misuse, users can choose from a library of curated, pre-built avatars. Custom avatar creation is subject to enterprise allowlisting and verification.

Google said all generated audio and video streams include imperceptible SynthID watermarks designed to make AI-generated content transparent and verifiable.

Gemini 3.8 Live with Live Avatar is available through US and EU endpoints, with provisioned throughput, enterprise compliance and data governance features.

Earlier, Qazinform News Agency reported that OpenAI had introduced GPT-6 Sol and GPT-6 Luna, two new models designed to bring the capabilities of its latest GPT-6 generation to a wider range of professional and everyday applications at a lower cost.