Computers are becoming more and more ubiquitous, moving from the desktop into our everyday life. Today’s challenge is to build a suitable visualization architecture for anthropomorphic conversational user interfaces which will run on different devices like laptops, PDAs and mobile phones. This new kind of interface will be adaptive to the current user, personal preferences, the history of the conversation, the device and the current context. Concrete implementations as a part of conversational interfaces are User-Interface Avatars, anthropomorphic representatives on the base of artificial 2D or 3D characters. The user can talk to an avatar on every device he is using. The avatar system is designed to exchange different graphical representations of the avatar easily. The existing system and ongoing work on optimization and renderers implementation are discussed.