The utilization of synthetic data is a prevalent method for obtaining substantial data points in a controlled environment, ensuring diversity and precise annotations. Synthetic data will give us accurate face landmark ground truth which is very difficult to annotate. However, a significant challenge persists in generalizing the performance of learnable models trained on synthetic data when evaluated against real data. Our proposed approach addresses this challenge by fitting a 3D morphable model to real data and synthetically morphing the reconstructed model. This creates facial data with updated landmark annotations in a controllable environment, offering diverse options for testing and evaluation. Using a mix of synthetic, real and augmented data should give an edge for training. In this session we will review methods which have been used successfully in other use case scenarios and evaluate their usability in OMS. Also show current status and results for OMS.