The technology herein involves use of 2D imposters to achieve seemingly 3D effects with high efficiency where plural objects such as animated characters move together such as when one character follows or carries another character. A common 2D imposter or animated sprite is used to image and animate the plural objects in 2D. When the plural objects are separated in space, each object can be represented using its respective 3D model. However, when the plural objects contact one another, occupy at least part of the same space, or are very close to one other (e.g., as would arise in a situation when the plural objects are moving together in tandem), the animation system switches from using plural respective 3D models to using a common 2D model to represent the plural objects. Such use of a common 2D model can be restricted in some implementations to situations where the user's viewpoint can be restricted to be at least approximately perpendicular to the plane of 2D model, or the 2D surface on which the combined image is texture mapped can be oriented in response to the current virtual camera position.