The systems and techniques disclosed herein provide tutorials for drawing three dimensional objects with accurate proportions and perspective. A user is able to select an object and a viewpoint to automatically generate a tutorial. Regardless of the object and viewpoint, an easy-to-use tutorial is produced that guides the user to draw the object with accurate proportions and perspective. Given a segmented 3D model of the object and a camera viewpoint, a sequence of steps for constructing the scaffold is determined. The sequence of steps is based on an intelligent selection of primitives and inter-primitive anchorings that provides an order for drawing the primitives and makes the scaffold easy to construct. The primitives and inter-primitive anchorings are selected from a rich set of possibilities that allow for some inaccuracies to reduce the difficulty of the tutorial. The primitives and inter-primitive anchoring are selected to balance the difficulty and the potential inaccuracy.