Inpainting

Inpainting lets you fill only a masked portion of your input image whilst keeping areas outside the mask consistent.

Inpainting works by specifying a mask that the AI will stay within whilst generating an image using your prompts. Everything outside of the mask will be kept consistent. At the moment, inpainting uses actor layers to help generate the mask so you will need to create an actor layer and add actors to it that you want to show up in the mask. In the example above, a sphere static mesh is added to an actor layer to act as a round mask that will limit inpainting to the cat's face.

To use inpainting, load up any image pipeline preset with Inpaint in its name to set up the required models, pipelines and layers. A good starting point is the SD1-5_Inpaint pipeline preset.

Expand the Pipeline Stages->Stage Layers->Layers property and find the layer that has a StencilLayerProcessor. Expand the Processor Options property and pick your actor layer from the dropdown to let the stencil layer processor know which actor layer we'll be using as a mask.