Gone are the days when the quality of machine learning models was limited by insufficient test and training data. DocumentCreator permits to generate unlimited amounts of semantically fully annotated document variations based on just a few original input specimens. DocumentCreator is particularly helpful in cases where documents for test or training data are few in number or cannot be used to train machine learning models because of data confidentiality, copyright issues or insufficient or inaccurate target value annotations. DocumentCreator is available as on-premise solution or as a software as a service (SaaS).