Threshold

Overview

Converts an image into a black-and-white mask of selected and unselected areas. Use it to extract only bright, nontransparent, or strongly red areas.

Raising threshold narrows the white range; lowering it widens the range.

Usage tips

• The shortest mask routes are noise_generate → threshold for random spots and original image → threshold(Alpha) for a silhouette.
• The go-to destinations are mask inputs on blend_by_mask, erase_by_mask, and mask-family nodes.
• Start with the default OKLCH-L, perceived brightness, for selecting bright areas. Switch to RGB/HSV only when extracting a specific color component. For a circular hue range that crosses red, use a color-selection node such as color_select_replace.
• Turn alpha_from_mask ON to make the black side transparent while creating the mask, producing a transparent-background black/white mask. It does not retain the original image colors.
• ignore_fully_transparent keeps fully transparent input pixels unselected regardless of hidden RGB. Graphs migrated from v1 retain the former false setting for compatibility.

Common pitfalls

• The boundary flickers or looks unstable: values near the threshold are varying. Move threshold by ±0.1 to find a stable point, or smooth the distribution with a light gaussian_blur upstream.
• All intermediate tones disappear: this node intentionally binarizes. Use posterize or gradient_map(Steps) to retain multiple stages.
• Black and white are reversed: simply turn invert ON; no upstream change is needed.

Examples

• Use Alpha to make a mask where only opaque pixels are white.
• Use OKLCH-L to extract only luminous areas.
• Turn invert on to select only shadows or dark areas.

Related nodes

• blend_by_mask / erase_by_mask — primary destinations for the mask
• noise_generate — source for a random mask
• posterize — reduce to several stages instead of two
• color_select_replace — select and replace a specific color