Inpainting Mask

An inpainting mask is the input that drives AI object removal: a map dividing the image into 'remove and rebuild this' versus 'keep this and use it as context.' When you brush over an object in Magic Eraser, you are painting the inpainting mask. The model then treats everything inside the mask as unknown and regenerates it from the visible pixels outside the mask, matching texture, color, lighting, and structure across the boundary. The quality of a removal depends heavily on the mask: too tight and the object's edge, shadow, or reflection leaks through; too loose and the model rebuilds more area than necessary, risking artifacts where it had ample real detail to keep. Good practice is to include the object plus a small margin — and any attached shadow or glow — so the reconstruction has a clean boundary to blend into. Auto-detection generates the mask for you (tap an object, the model proposes the mask); manual brushing gives precise control on tricky edges. The inpainting mask is the bridge between what the user wants gone and what the AI Fill model actually reconstructs.