Currently auspicious clouds are mainly created by hand. Most of previous research works focus on realistic cloud and smoke simulation， and little effort has been paid to the cloud patterns with traditional Chinese style. A data-driven and sketch-based cloud generation model with traditional Chinese style is presented. First， the feature curves are extracted from the hand-painted auspicious cloud pattern samples and normalized to form a feature library. The user draws a simple two-dimensional sketch to represent the skeleton curves of the cloud head， cloud body and cloud tail respectively. The algorithm performs non-uniform automatic division of the cloud body sketch trajectory to obtain a generated line segment set and then find the suitable curve depending on every line segment in the feature library and mount it. The cloud tail part is procedurally generated by panning and scaling the cloud tail skeleton curve. Experimental results show that the proposed model is able to generate a wide variety of auspicious cloud patterns with hand-painted style features with little interaction and parameter control.