Do I need to (re)train every time the real data changes?

When our self-learning system has trained itself to create synthetic data, it only needs to run periodically to keep up with the changing characteristics of your real data. To save time, you can do this incrementally and over different time intervals, depending on your use case.