I'm working on the Synthetic Data Vault, a set of libraries to generate realistic synthetic data. We have an option to use GAN models but we've found that Gaussian & other copula models work great because they're faster and more user configurable.

https://github.com/sdv-dev/SDV