I propose a two-factor variance model that extracts the common dynamics in stock variances by imposing factor structures directly on the variance series. The two factors can be identified independently as a market variance (MV) factor and a variance residual (VR) factor, respectively governing the short-term and long-term dynamics of individual stock variances. I show that the MV factor resembles the shape of the index variance and carries a positive premium, which has an excess spillover effect on individual stock variances. On the contrary, the VR factor serves to compensate the excess spillover by having a negative premium. The differences in the factor premia and factor memories suggest that an option portfolio with long positions on long-term individual stock straddles and short positions on short-term index straddles generates significant positive returns. The strategy can be further enhanced by choosing straddles according to their model predicted variance risk premia. The advantage of the model-based strategy is demonstrated in both in- and out-of-sample analysis.