The university course timetabling problem is hard and time-consuming to solve. Profits from full automatisation of this process can be invaluable. This paper describes architecture and operation of two automatic timetabling systems. Both are based on evolutionary algorithms, with specialised genetic operators and penalty-based evaluation function. The paper covers two problem variations (theorethical and real-world), with different sets of constraints and different representations. Moreover, specification of both solutions and a proposal of hybrid system architecture is included.