StableDrag is a point-based image editing framework designed to address the issues of inaccurate point tracking and incomplete motion supervision in existing drag-and-drop methods. It employs a discriminative point tracking method and a confidence-based latent enhancement strategy. The former accurately localizes updated handle points, improving stability for long-distance operations; the latter ensures that the quality of the optimized latent representation is as high as possible throughout all operational steps. The framework instantiates two image editing models, StableDrag-GAN and StableDrag-Diff, which demonstrate more stable drag performance through extensive qualitative experiments and quantitative evaluations on the DragBench."