We review the application of non-Abelian discrete groups to Tri-Bimaximal (TB) neutrino mixing, which is supported by experiment as a possible good first approximation to the data. After summarizing the motivation and the formalism, we discuss specific models, mainly those based on A4 but also on other finite groups, and their phenomenological implications, including the extension to quarks. The recent measurements of θ13 favour versions of these models where a suitable mechanism leads to corrections to θ13 that can naturally be larger than those to θ12 and θ23. The virtues and the problems of TB mixing models are discussed, also in connection with lepton flavour violating processes, and the different approaches are compared.