Abstract: Devices and methods are disclosed which relate to improving the efficiency of text input by dynamically generating a soft input based upon current context information. In some embodiments the dynamic soft input may comprise a reduced set of input areas (e.g., keys), which may be sized and/or positioned according to their relative probability of being selected as the next user input. In addition, in some embodiments the probability that an input will be selected may be determined by comparing the current context (e.g., user inputs) to an input model, such as a language model.