Sketching captures spatial structure and shading intent directly, but existing sketch-to-3D pipelines have no way to use that intent: they turn drawn highlights and shadows into dark paint, locking the lighting into the texture. The paper proposes a system that instead takes a rough garment sketch annotated with a star for a highlight and a moon for a shadow, together with a short text prompt describing the texture, and produces a textured 3D mesh with an explicit directional light.
The key idea is to separate lighting from appearance. From the marked sketch, the system predicts the scene's directional light and, independently, a displacement field that bends the mesh surface so the moon-marked regions fall into shadow under that light, rather than being painted dark. The star and moon symbols were the most frequent proposals in an elicitation study with art students, making them intuitive annotations for nonexpert users.
The result is a scene that can be relit and edited after generation, instead of a static texture. This moves sketch-based 3D creation closer to a workflow where the designer's original shading intent remains editable, rather than being baked into the surface.