The paper proposes a simple yet elegant change to make DP GANs significantly more accurate and presents the finding clearly. All reviewers agree that the technical novelty is not especially big, but as the method appears practically significant, the paper should be accepted.