Latest improvement in machine learning have significantly improved the performance of facial identity detection systems. The performance and reliability of these models depend heavily on the large amount and good quality training data. However, the collection of annotated large data-sets does not scale well and the control over the quality of the data decreases with the size of the data-set. In this talk, I will present how synthetically generated data can be used to decrease the number of real-world images needed for training deep face recognition systems. I will discuss in details about 3D morphable face model for generation of identities and samples for these synthetic identities using pose, illumination and background as parameters. I will showed the results of facial identity detection on real and synthetic data using state-of-the-art face model (arcface). In the end, I will discuss about real-to-virtual performance gap and how synthetic data will helpful to reduce that gap.