We propose a novel method for GPU-based rendering animated smoke in pen-and-ink style.We use physical-based simulation data generated by a GPU-based smoke simulation algorithm as input for our NPR system in order to ensure that the animation is in accordance with real smoke’s behavior in reality. Our stylized smoke is constructed by two components, silhouettes and shaded interior smoke regions. While the silhouettes are rendered by several strokes, the visible interior regions of the smoke can be shaded by NPR color maps or strokes or nothing, depending on artistic style. In the case of using strokes, we utilize maximum curvature direction values and vectors as a clue to deposit strokes which are rendered by texturing. Our algorithm is the first work of pen-and-ink smoke rendering. Since our stroke rendering is surface-based, the rendered NPR smoke convey well shape of the simulated smoke. Due to its GPU-based implementation, our algorithm could generate frames in interactive rate. In addition to pen-and-ink style, our algorithms also could provide various NPR styles easily by changing stroke textures, stroke directions or using cartoon coloring instead of strokes.