Google’s new AI avatars can now talk and show emotion

The new feature allows businesses to use an animated character to handle tasks like customer service or digital walkthroughs. These avatars are designed to react as you speak, matching their lip movements and expressions to what the AI is saying at that exact moment. They are built to work across 97 different languages, adjusting how they move their mouths to match the sounds of each language as they switch between them. Organizations can either choose from a library of preset faces or create custom avatars that match their specific brand identity. To keep things transparent, every video generated by these avatars includes an invisible digital signal called SynthID, which marks the output as AI-created so it can be identified later.

When you talk to a standard AI, it usually just produces text or audio. The Live Avatar goes a step further by combining three separate technologies at once. First, it uses the underlying intelligence—the model—to understand your words and decide how to respond. Second, it uses a generator that creates video in real-time, effectively drawing the character’s face and movements as it speaks. Finally, it uses a synchronization system that forces the digital mouth to move in rhythm with the specific sounds of the words being spoken. To prevent the conversation from lagging, the AI performs these tasks in the background. It can also look up information from a company database while it is still talking to you, ensuring the chat feels like a natural conversation rather than a long, clunky wait while the computer thinks.

This technology represents a shift from AI as a static tool to AI as a conversational partner. By adding visual cues like expressions, companies are betting that digital interactions will feel less like filling out a form and more like speaking to a human representative. However, as AI becomes more capable of mimicking human appearance, the risk of confusion increases. By including digital watermarks, Google is acknowledging that it needs to provide a way to verify what is real and what is synthetic. The challenge for the future will be whether these avatars truly make experiences better for us, or if they simply make digital interactions more persuasive and harder to distinguish from reality.