An attached image can provide a starting point or guidance for generation.
Choose a mode according to the change you want:
generates from a noisy version of your picture.
limits that change to a painted area while retaining the original content as a starting point.
replaces a painted area using the prompt and surrounding image.
fills added space around your picture.
uses image references to guide Stable Cascade generation.
ControlNet uses a map of features such as outlines, depth, or pose to guide a new image.Model and worker support differ between modes. For masked work, use the supported Stable Diffusion 1.5, 2, or SDXL families and check the studio's compatibility warnings.