downup-sample-blur — Traditional AI Model
Blur on the cheap: shrink the image, blur the small copy, scale it back up. Faster than full-resolution Gaussian and coarser at the edges, which is fine for a quick backdrop or a test frame. CPU, sub-second.
Details
- CategoryTraditional
- Year-1
- License
Inputs & Outputs
- ImageInputimage
Source image to blur
- ImageOutputimage
Blurred output image
Parameters
- kernel_sizeintegerdefault 5range 3–15
Gaussian kernel size (3-15)
Size of the Gaussian kernel applied after downsampling. Must be odd. Larger values produce a softer post-upsample result. Default 5 is a good baseline; bump to 9-15 only when you want extra smoothing on top of the downsampling effect.
- color_multiplierfloatdefault 1range 1–30
Brightness multiplier (1-30)
Multiplies output pixel values to brighten the result. 1.0 keeps original brightness; values above ~3 risk clipping highlights to pure white. Useful when the down/upsample chain darkens the image. Default 1.0.
- scale_factorintegerdefault 10range 2–30
Downscale ratio (2-30)
How aggressively to downscale before blurring. Higher values produce much blurrier output much faster, but lose more detail. Default 10 yields a strong but recognizable blur. Use 2-3 for subtle softening, 15-30 for near-abstract backgrounds.
Tags
Alternatives in Blur & Effects
- Image Blur (Simple)
Single-slider blur (1-10). Fastest blur option.
- Image Blur (Standard)
Standard Gaussian blur with iteration control.
- Lens Blur
Circular disc-kernel blur for camera-bokeh look.
- Selective Blur
Blur only a drawn region (rect/circle/polygon).