Microsoft tests human-like animated avatars for Copilot AI in US, UK, Canada

Microsoft is testing a new feature called Copilot Portraits in the US, UK, and Canada, which introduces 40 AI-generated avatars that provide a more human-like interaction by displaying facial expressions, head movements, and lip-syncing during real-time voice conversations. Powered by the advanced VASA-1 technology, these avatars are created from a single image without complex 3D modeling and are deliberately stylized to balance realism with user safety. This initiative responds to user feedback expressing a preference for speaking to a face rather than a disembodied voice, aiming to make AI interactions feel more immersive and personal without misleading users into thinking the AI is human. Microsoft emphasizes privacy and security, limiting the feature to users 18 and older and including usage limits and clear AI indicators on screen. The move marks a shift from previous, more cartoonish representations toward more human-like avatars, contrasting with competitors like X's 3D avatars, while adopting a cautious rollout to avoid harmful interactions. Overall, Microsoft is advancing its AI assistant Copilot to feel more approachable and natural through expressive, customizable avatars, supporting a more conversational AI experience.