Organizations across DoD and DHS use training to develop the visual search skills necessary to compare facial features of individuals to identification documents (ID) photos and determine if the individual is an impostor.  This task presents a challenge, as individuals may differ in age and appearance compared to the provided ID photo. Training to effectively gain these visual search skills is typically accomplished through instructor-led presentations in a classroom setting, with instructors describing the techniques and highlighting the critical visual cues (e.g., facial features) needed to perform the task. However, few training platforms currently have a means to visualize performance during the task and obtain objective feedback of how well they visually interrogated critical cues. Creating training platforms to address these issues is challenging due to the lack of images that are (1) photorealistic, (2) appropriately different across ID and individual photos, and (3) publicly available for use in a training platform.  Advances in deep generative methods, Artificial Intelligence (AI) algorithms such as Nvidia’s StyleGAN2, allow for the creation of photorealistic artificial face images that can (1) avoid privacy issues and (2) allow for artificial aging and other methods to make the images look more appropriately different.  Advantages to using deep generative methods artificial face images also includes the ability to update an image database on a continual basis and the ability to more closely control the demographics of the images used.  However, the act of creating and changing face images results in challenges associated with validating that pairs of face the “same person” or an “impostor”.  The current paper examines the issues associated with using existing deep generative methods to create  a database of facial images appropriate for impostor detection training, the methodology utilized to create a set of training images, and the process used to validate the images.