Stellar atmospheres represent unique windows for understandingstellar, galactic and cosmic evolution by being responsible forthe emission of stellar spectra. Much progress has been made overthe years in modelling stellar atmospheres but still the modellingefforts are hampered by various, often questionable, assumptionsand approximations. This review describes promising avenues forimproving the realism of stellar model atmospheres for hot(spectral types O, B, A), cool (F, G, K)and very cool (M and later) stars, respectively, in the coming decade.A common theme will be time-dependent 3D hydrodynamical calculationswith a detailed non-LTE treatment of the radiative transfer.It is argued that this is fully within the realm of possibilityon this time-scale and indeed will be necessary to complementthe expected advances on the observational side.