This repository contains the full artifact package for our USENIX Security '25 paper, SoK: Can Synthetic Images Replace Real Data? A Survey of Utility and Privacy of Synthetic Image Generation. It includes all necessary components to reproduce the experimental findings and results presented in the manuscript. This artifact package contains: Source Code: All Python (.py) and shell (.sh) scripts required to train all generative models and downstream classifiers from scratch, generate synthetic data, and execute all three privacy attacks (Attack 1, Attack 2, and Attack 3/LiRA). Processed Datasets: A dataset.zip file which contains the exact processed CSV files (train.csv, test.csv, and all data splits required for our Membership Inference Attacks) and corresponding image subsets for all three datasets: CelebA, Fitzpatrick17k, and CheXpert. Documentation: A comprehensive README.md file with step-by-step instructions for setting up the computational environment (requirements.txt), preparing the data, and running the experimental pipelines to reproduce our results. How to Use: To get started, please download the artifact package, unzip the dataset.zip file, and follow the instructions in the README.md file.