A mask is a defined region within an image or video frame that controls where an effect, edit, or generation applies, protecting areas outside the mask from being affected. Masks work like stencils - they specify which parts of an image are active and which are shielded, allowing precise, localized changes without affecting the entire frame.

In traditional compositing, masks are drawn around subjects to separate them from backgrounds, enabling independent processing of each layer. In color grading, masks isolate regions for targeted exposure or color adjustments - brightening a face without affecting the background, for example. In AI generation, masking has become a central workflow tool through inpainting, where a mask defines the area the model should regenerate while leaving the surrounding image untouched. This allows creators to fix specific problem areas in an otherwise good generation, replace elements selectively, or extend an image by masking newly added canvas space for the model to fill. Masks can be drawn manually, generated automatically from subject detection, or derived from depth or segmentation data.

Masking is one of the most powerful precision tools available in AI image editing workflows. Being able to direct regeneration to specific areas of an image - a face, a background, an unwanted element - without disturbing the rest of a carefully crafted composition makes iterative refinement significantly more efficient than regenerating the full image from scratch every time a single area needs adjustment.