A generative modeling technique that learns to transform random noise into realistic data by following learned flow paths.