Blend modes¶
Each layer (and group) composites using a blend mode chosen from the per-layer dropdown. PixelArt Creator ships twelve modes — Normal plus the eleven non-normal separable modes from the W3C Compositing and Blending Level 1 specification (the same set and results as Photoshop / Krita / SVG).
Blending is done on straight (non-premultiplied) alpha in a normalised floating-point working space, then written back to 8-bit RGBA. Normal is straight source-over alpha compositing.
The twelve modes¶
| Mode | Effect |
|---|---|
| Normal | Straight alpha source-over. A fully opaque layer replaces what is below; a fully transparent layer leaves it unchanged. |
| Multiply | Multiplies the layer with what is below — always darkens. White is neutral; black yields black. |
| Screen | Inverse of multiply — always lightens. Black is neutral; white yields white. |
| Overlay | Multiply in the dark tones, screen in the light tones — boosts contrast (relative to what is below). |
| Darken | Keeps the darker of the two channel values. |
| Lighten | Keeps the lighter of the two channel values. |
| Colour Dodge | Brightens what is below to reflect the layer; strong lightening. |
| Colour Burn | Darkens what is below to reflect the layer; strong darkening. |
| Hard Light | Overlay with the layer and backdrop roles swapped — a harsh spotlight. |
| Soft Light | A gentler hard light — soft dodge / burn depending on the layer value. |
| Difference | Absolute difference of the two values — equal colours cancel to black. |
| Exclusion | Like difference but lower contrast in the mid-tones. |
Opacity and blend mode combine
A layer's opacity scales its contribution for every blend mode, not only Normal. A hidden layer contributes nothing regardless of mode.
How the stack is composited¶
Layers are composited bottom-to-top over a running result: each visible layer's pixels are blended over what is below using that layer's blend mode and opacity (and its mask, if any). A stack of Normal layers is exactly the same as folding straight alpha source-over from the bottom up. Compositing never mutates the source layer buffers — it is non-destructive.
A single edit recomposites only the affected region, so painting on a large canvas stays responsive (see the performance notes below).
Performance¶
On an 8K canvas (7680 × 4320) a single-pixel paint recomposites its small dirty region in about 1 ms, comfortably inside the 16 ms / 60 fps frame budget. Some heavier interactions at low zoom — dragging the opacity slider, or an attribute change that repaints a large viewport across many layers — are currently CPU-bound and can exceed the budget; full GPU-shader compositing is a planned later enhancement.
See the changelog Known limitations for details.