Devices and methods are disclosed which relate to improving the efficiency of text input by dynamically generating a visually assistive virtual keyboard. Exemplary variations display a soft keyboard on a touchscreen of a text-entry device. The touchscreen works with the soft keyboard as a form of text input. Keyboard logic on the text-entry device is programmed to change the visual appearance of each key within the soft keyboard based on the prior entry. The keyboard logic assigns a prediction value to each key based on a statistical probability that the key will be entered next. The touchscreen displays a visually enhanced keyboard based on these prediction values. Enhancements include resizing keys relative to their prediction value, rearranging the keys a distance from the previous key entered inverse to its prediction value, etc.