I think it would go the other way where you use coverage to guide reinforcement. Crank the temperature up to increase variation and you would probably produce a model that could approximate the file format you were targeting.