Try Pro FREE for 2 weeksGet Started

image-apply-alpha — Masking AI Model

Apply Mask as Alpha turns an image and a black-and-white mask into a transparent cutout: white keeps, black becomes transparent, grey is partly see-through. Image Multiply can only paint the dropped area black; this leaves it clear, so the result drops onto any background. If the image is already a cutout (BiRefNet), the two are combined, so BiRefNet's soft hair edge and a SAM mask that isolates one person work together. A light feather softens the stair-stepped edge of segmentation masks. Output is always PNG. CPU, instant.

Details

  • CategoryMasking
  • Year-1
  • License

Inputs & Outputs

ImageInputimage

Image to cut out (an existing alpha is kept and combined)

MaskInputimage

Mask: white = keep, black = transparent

ImageOutputimage

Transparent cutout (PNG)

Parameters

featherintegerdefault 1range 0–32

Soften the mask edge (px)

Gaussian blur radius applied to the mask before it becomes alpha, in pixels of the image. Default 1 smooths the stair-stepped edge a segmentation mask leaves; 0 keeps the mask exactly as given; 3-6 gives a visibly soft edge. Applied to the mask only, so an alpha the image already has (BiRefNet) keeps its own edge.

invertbooleandefault false

Invert the mask (keep the black area)

When true, black keeps and white becomes transparent — e.g. to cut the subject OUT of the image and keep the background. Default false.

fitstringdefault coveroptions: cover, contain, stretchadvanced

How to fit a differently-sized mask to the image.

When the mask's resolution differs from the image's, how the mask is fit to the image (the output is always the image's size): 'cover' (default) scales to fill and center-crops, 'contain' scales to fit and pads with transparent, 'stretch' resizes to the exact size (use it when the mask came from a resized copy of a different aspect).

Tags

  • transparency
  • alpha
  • cutout
  • background-removal
  • mask

Alternatives in Image Masking

All AI models