Genetic mark-recapture (GMR) is a statistical technique used in estimating population size in ecology. By combining the relative abundance of species based on genetic data with population counts obtained for some of the species, GMR allows the estimation of the total population size and the contributions of each species. The current method uses a Lincoln-Petersen estimator and provides an asymptotically unbiased estimate for the total population size. However, the variance estimator does not account for all the uncertainty in the sampling process of the genetics data. As a result, this approach can suffer from a significantly underestimated variance, especially when the genetic sample size is small. In this work, we propose two novel Bayesian GMR models and a modified variance estimator based on the original method to address this issue. The effectiveness of the new methods is investigated via simulation studies and a real data analysis is conducted in the study.