This DragGAN (OpenGVLab) project is an unofficial open-source implementation of the Drag Your GAN paper, bringing interactive point-based image editing to anyone’s machine. Published by OpenGVLab on GitHub, it has gathered more than 4,900 stars. Note that this is a distinct project from other DragGAN-related products listed on Tooloso, and covers this specific implementation.
DragGAN lets you edit a generated image by simply dragging points on it: move a handle point toward a target and the image content follows in a semantically sensible way, opening a mouth, changing a pose or reshaping an object. The method works on the generative image manifold of a GAN rather than painting pixels directly. This implementation ships the full pipeline described by its authors, with model code released openly, a Gradio interface for interactive use, a Colab demo for quick trials, and local deployment support on Windows, macOS and Linux. The repository was last updated in 2023, so treat it as a complete research release rather than an actively evolving product.
Because the code, models and demos are fully public, the repository also serves as a practical teaching resource for how latent-space editing works in practice. Its release history documents steady bug fixes and speed improvements after the original publication, and the companion Awesome-DragGAN collection gathers papers and derivative projects for readers who want to explore the area further.
The implementation is available as free open-source code on GitHub and PyPI. You only need your own hardware, ideally a GPU, to run the models locally.
DragGAN (OpenGVLab) is for researchers, students and creative developers who want to experiment with intuitive, drag-based generative image editing. It is a good reference for anyone studying controllable generation or building interactive editing prototypes on top of GAN latent spaces.