By using the well known Selberg integral we define a bivariate beta distribution. Unlike other bivariate beta distributions which have support on the simplex {(x,y):x>0,y>0,x+y<1} this distribution has support on {(x,y):0<x<1,0<y<1}. Several properties such as marginal and conditional distributions, joint moments, coefficient of correlation, entropies, Fisher information matrix and estimation of parameters have been studied.