BGCut

Edge refinement guide

Fix difficult cutout edges without starting over

The fastest correction is the one matched to the type of error. Missing objects, leftover background, thin halos, and soft fur each want a different tool — not the same brush.

Dog photograph before background removal edge refinement
Original subject and background
Dog cutout with transparent background after BGCut processing
Automatic result ready for inspection

Know your enemy: the four edge problems

Move the comparison slider across transparent, white, and gray previews and decide which of these you are looking at before touching a tool.

Missing subject

An ear, wheel, sleeve, or petal is gone. Fix with smart restore — click or stroke inside the missing part.

Leftover background

A patch survives between legs, inside a handle, under an arm. Fix with smart erase inside the unwanted region.

Halo / fringe

A thin bright or dark ring from the old background. Fix with a manual erase brush along the edge.

Soft material

Fur, hair, steam, defocus. Often not an error at all — keep partial transparency; use a soft brush only where color spill is real.

Repair from large to small

  1. Get the silhouette right at normal zoom

    Shape errors matter more than pixels. Restore or erase the big regions first — scanning every pixel at maximum zoom is slow and produces unnaturally hard edges.

  2. One smart stroke, then verify

    After each smart operation, hold the original comparison and walk the boundary of what changed. If it grabbed too much, undo once and retry smaller — never stack prompts on a bad selection.

  3. Manual brushes for the last few pixels

    Isolated specks, narrow halos, details no model can guess. Brush slightly smaller than the correction; harder for product edges, softer for fur and fabric.

  4. Stop at destination quality

    A one-pixel defect at 800% zoom is invisible in a 160-pixel thumbnail; a missing ear never is. Judge the result at the size and background where it will be published.

Before you export

  • The overall silhouette was correct before pixel-level cleanup began.
  • Every smart operation was verified immediately and undone if it touched unrelated areas.
  • Manual brush hardness matches the material at each edge.
  • The result was checked on more than one background at delivery size.
  • The transparent master is downloaded before any layout or resizing work.
Try BGCut