Making facial computer animation look convincing has proven to be a difficult task. A common limitation of physicallymodeled faces [Essa 1995, Essa et al. 1994, Waters 1990, Takeuchi & Nagao 1993, Waite 1989] is that the meaning of their expressions is often vague. An ideal solution to this would be to exaggerate facial expression, but within a physical modeling framework this may look unconvincing or awkward. An alternative is what might be called a ÒcaricatureÓ approach [Thrisson 1993, 1994a, 1994b, Britton 1991, Laurel 1990] where details in the face are minimized and the important features therefore exaggerated (see Hamm [1967] for an excellent discussion on cartooning the head and face). In this fashion, Brennan [1985] created a system that could automatically generate caricature line-drawings of real people from examples that had been entered by hand. Librande [1992] describes a system called Xspace that can generate hundreds of artistically acceptable two-dimensional drawings from a small example base. Simplified faces seem like a very attractive alternative to physical modeling for animating interface agents, both in terms of computational cost and expressive power. This report describes a system called ToonFace that uses a simple scheme for generating effective facial animation along these lines (Figure 1). It differs from prior efforts primarily in its simplicity and its way of representing facial features. The following section is a more elaborate discussion of the motivation and goals behind this work. Section 3 describes the particulars of drawing and animation routines, section 4 gives a quick tutorial of the two parts of ToonFace, the Editor and the Animator. A short user guide follows in section 5 gives a short comparison between ToonFace and the Facial Action Coding Scheme (FACS, Ekman & Friesen 1978). Lastly, current applications and future enhancements are described in section 6.