Machine learning enjoys a lot of attention in the recent years as a straightforward solution for complex challenges. This is especially true for machine vision in application such as visual surface inspection or modelling of material characteristics. What often becomes apparent is the drastic lack of data necessary to train such models in a way which will cause them to perform well when deployed. Therefore, the machine learning community is turning its attention to synthetic data as a way to make up for the missing real data. However, there we have a new set of challenges which need to be overcame. One of such challenges is material representation in technical imaging. We tackle this challenge by using combination of computer graphics, physics and mathematical models developed to recreate material properties. This approach gives us a possibility to parametrically generate arbitrary amounts of synthetic data with complete control over its variability, content and balance.