With Syn4Syn we are synthesizing 3D-data sets true-to-life on the basis of measured real world data with the aim to train 3D image processing using deep neural networks. Thus, we want to establish possibilities to automatically and robustly evaluate arbitrary image stacks of 3D cell cultures generated in pharmacy or medicine. Imaging of 3D cell cultures suffer especially from optical effects (point spreading, bleaching, vanishing fluorescence in depth etc.) and biophysical processes (surface energies, migration movements etc.) which lead to inhomogeneities which need to be physically modelled. The physical models are to be fed into the training of deep neural networks, to improve robustness even if only few training data are used and thus create realistic synthetic images (containing ground truth). To support applicability to researchers not being familiar with computer science, all algorithms are encapsulated in graphical user interfaces and comprehensively documented. The algorithms will be available in GIT-Repositories, data will be stored on cloud-platforms.