This paper describes a system which generates variations on theme music fitting to story scenes represented by texts and/or pictures. Inputs to the present system are original theme music and numerical information on given story scenes. The present system varies melodies, tempos, tones, tonalities, and accompaniments of given theme music based on impressions of story scenes. Genetic algorithms (GAs) using modular neural network (MNN) models as fitness functions are applied to music generation in order to reflect user's feeling of music and stories. The present system adjusts MNN models for each user on line. This paper also describes the evaluation experiments to confirm whether the generated variations on theme music reflect impressions of story scenes appropriately or not.