DiTGaze is a calibration-efficient gaze augmentation system for few-shot person-specific gaze estimation. An operator adopting it should change the generator, not the downstream estimator: replace the convolutional rendering stage of a two-stream 3D Gaussian Splatting backbone with a geometry-conditioned latent Diffusion Transformer (DiT).