Bayesian predictive density functions of the median of a set of odd and even number of future order statistics are obtained when the observations (informative and future) are assumed to follow a finite mixture of components of general form and type 1 censoring is imposed on the informative sample. The prior belief of the experimenter is measured by a general class of distributions which includes most priors used in literature. A mixture of two Weibull components model is given as an application. A numerical example presents Bayesian prediction bounds of the future median of observations based on a finite mixture of two exponential components.