Variational inference for Bayesian Gaussian mixture model: compatible version for fixed known number of components?

Question

I have found examples of R code for Gaussian mixture model decomposition using variational inference with unknown number of components. In the variational inference, the iterations modify the number of components during the multiple iterations. It makes sense for classic clustering with no info on the number of clusters. However, can we address a seemingly simple problem with fixed known number of components? In other words, is there a possibility to use variational inference when the number k of mixture components is known and fixed? In a classic univariate MCMC bayesian model (for example with JAGS and R Bayesmix package) I can specify a fixed k of components but avoiding MCMC burden and its computation time would be great.