We propose a dependent frequency-severity model using a Gaussian copula. The copula links a latent variable of waiting time for the second claim with the claim severity. By assuming a log-normal distributed claim severity, we can analyze the effect of claim counts on the conditional expectation of severity. We propose a Monte Carlo simulation algorithm to simulate the predictive distribution of the aggregated claims amount. In an empirical example, we compare the proposed method with the conditional modeling by Garrido et al. (2016) and the mixed copula modeling by Czado et al. (2012).