DragGAN

DragGAN is a research method (SIGGRAPH 2023, from Max Planck Institute for Informatics and collaborators) for interactive point-based image manipulation: you place handle points on a GAN-generated image and drag them to target positions to precisely control pose, shape, expression, and layout across categories like animals, humans, cars, and landscapes. It's aimed at computer-vision and graphics researchers and creative technologists rather than commercial users, and it can also edit real photos via GAN inversion. What sets it apart is its combination of feature-based motion supervision with GAN-feature point tracking, giving pixel-level 'drag' control without annotated training data or a prior 3D model.