Choose a blur construction

The soft style averages a horizontal neighbourhood and then averages the result vertically. Directional styles perform only one pass. Every output pixel is therefore an explicit mean of a rectangular line or area, with no semantic subject detection and no hidden depth estimate.

Blur direction

Two-axis soft, horizontal smear or vertical smear.

Radius

How far the average reaches on each active axis.

A uniform radius treats foreground, subject and distant background equally. It cannot reproduce optical depth of field. Directional blur can also make bright points stretch into bands that dominate the frame, and a large two-axis radius can erase the exact subject separation the image needed.

The preview, draft and export all run the same kernel. Parameters measured in pixels are scaled with the source, so a chosen tile, band or radius does not become unexpectedly finer in the downloaded file.

Is soft blur Gaussian?

No. It is a separable box average: horizontal then vertical. The algorithm is stated because its edge response differs from a Gaussian even when both look soft.

What do horizontal and vertical modes simulate?

A one-axis average produces a directional smear. It can suggest motion, but it is not deconvolution and does not estimate what a camera actually did.

Why does a large radius darken or lighten edges?

Samples outside the frame are clamped to the nearest edge pixel. A strong border colour therefore contributes repeatedly near the boundary.

• Edge detectA Sobel luminance gradient map with threshold and inversion.
• PaintingSmoothed regions, reduced colour steps and measured contour weight.
• PixelateSquare or hexagonal blocks, each one flat, with optional palette quantisation.
• The effects labAll six effects, chosen from your own photograph rather than from a list of names.

Rendered on your computer, not on a server.