Microsoft is enhancing its Copilot AI assistant by introducing a more engaging feature: animated avatars. The experimental “Portraits” feature is now available in the United States, the United Kingdom, and Canada. This innovative addition offers users a choice of 40 stylized human avatars that exhibit natural expressions during real-time voice interactions.
Microsoft AI CEO Mustafa Suleyman announced the feature on X, explaining that the development stemmed from feedback indicating some users would prefer conversing with a face when using voice. The option to select a portrait and pair it with a voice aims to cultivate a more organic conversational experience compared to traditional text interactions.
This initiative marks Microsoft’s ongoing efforts to mold Copilot into a character that users find relatable. The company had previously unveiled a feature called “Copilot Appearances” in July, which differed in that it used cartoonish designs rather than human-like representations.
“This experiment utilizes VASA-1, a cutting-edge AI technology developed by Microsoft Research, which enables real-time visual AI-based conversations,” stated Microsoft on the Copilot Discord server. “The technology allows for the generation of lifelike facial expressions, head movements, and lip-syncing from a single image, eliminating the need for complex 3D models.”
This Portraits initiative follows another AI company, X, which recently introduced 3D avatars for its Grok chatbot, including a virtual companion designed with anime aesthetics. Microsoft is proceeding cautiously with the Portraits rollout, likely due to concerns over potential risks linked to user interactions with competing chatbot technologies, such as those under review by Character.AI.
“Each portrait is purposefully stylized rather than photorealistic, providing users with visual diversity while maintaining the trusted intelligence and safety features of Copilot,” Microsoft stated on their Copilot blog. Access is limited to a “small group of users” aged 18 and over, and includes daily and session time limits along with “clear indicators that users are interacting with AI technology.”