Skip to content

Filter Blur

Xaymar edited this page Jun 9, 2021 · 16 revisions

Blur Filter Windows Linux MacOS Unstable

With blur you can hide information, create complex glow effects, or just enhance the background of your transparent webcam overlay! Whether you choose to put more focus on yourself by blurring the game behind the webcam, or if you just decide to hide a small area on the screen - the choice is yours!

Guides

Examples

Settings

Type

Selects the kernel type to use for filtering, each has a unique effect on the output image.

Subtype

Changes the filtering mode for this kernel, not all modes may be supported by all kernels.

Size

Unit-less size of the blur. Unless otherwise stated, assume no unit applies.

Angle

Angle of the rotational blur, in degrees. Does nothing if size is too small.

Center

Center of the rotational or zoom blur, in percent.

Step Scaling

Enabling step scaling allows you to control how much the sample point moves for each kernel step. Higher values than 100 can be used to simulate a larger blur size than you've actually selected. Feature availability depends on kernel type.

Apply a Mask

Enables masking which does nothing more than blend between a fully blurred image and the original image. It can either work based on a rectangular region, an image, or another source.

Kernels (Types, Subtypes)

Box

Named after its distinctive Box like shape, it is very simple to calculate but also does not look great. It supports all filter types.

Box Linear

A slightly more optimized version of Box, which sacrifices accuracy for a 50% performance gain. It only supports Area and Directional filter types.

Gaussian

A gaussian blur kernel where all weights are determined by a normalized gaussian bell curve. Supports Area, Directional, Rotation and Zoom filtering types. It is very expensive to use.

Dual-Filtering

A ~98% correct approximation of the Gaussian kernel. It is incredibly lightweight compared to the real thing, and can even run on mobile early mobile GPUs in real time. The size is set as pow(2, N), so it is very granular in control - however at the same time it is insanely more performant than Gaussian Blur at comparable sizes.

Clone this wiki locally