Generates photorealistic images and videos with perfect, automatic labels (bounding boxes, segmentation masks) rendered from virtual environments. This provides a scalable, cost-effective source of training data for applications where real data is difficult to acquire.