Google Launches Gemini 3.8 Live Avatar with Real-Time Video and Multilingual Lip-Sync
Google has expanded Gemini 3.8 Live with a Live Avatar feature that combines near real-time video generation with speech output. The capability supports precise lip-syncing and dynamic facial expressions across 97 languages for enterprise AI agents. This update advances human-AI interaction by giving conversational agents a realistic visual persona that can listen, observe, and speak fluidly. It enables enterprises to build more engaging digital customer service experiences without interrupting conversation flow when executing background operations. Live Avatar leverages asynchronous tool calling to execute backend tasks and retrieve data without pausing the active conversation, and it embeds invisible SynthID watermarks directly into audio and video outputs. Enterprise developers can also create custom avatars from reference images, though custom creation is currently restricted to whitelisted enterprise accounts.
## BACKGROUND
Asynchronous tool calling allows large language models (LLMs) to invoke external functions, such as API requests or database queries, in the background while continuing to generate responses or maintain active dialog without blocking execution. SynthID is Google DeepMind's watermarking technology that embeds imperceptible statistical markers directly into AI-generated media to assist in detecting AI-generated content and preventing misattribution.