Build a CSS filter visually. Drag sliders for blur, brightness, contrast, saturation, grayscale, sepia, invert, hue-rotate and opacity, add a drop shadow that follows transparent edges, and compare the original and filtered image side by side — on the sample or on your own photo. Copy the result as filter or backdrop-filter. It's free and runs entirely in your browser; your image never leaves your device.
| Function | No-effect value | Unit |
|---|---|---|
| blur() | 0 | px |
| brightness() | 1 | ratio |
| contrast() | 1 | ratio |
| saturate() | 1 | ratio |
| grayscale() | 0 | ratio (0–1) |
| sepia() | 0 | ratio (0–1) |
| invert() | 0 | ratio (0–1) |
| hue-rotate() | 0deg | deg |
| opacity() | 1 | ratio (0–1) |
filter changes the element itself. backdrop-filter changes what's visible behind the element — it needs a translucent background to show.
drop-shadow follows the shape of transparent images and SVGs. box-shadow always draws a rectangle.
Yes, they're applied left to right. This tool writes them in a consistent order.
Most are cheap and GPU-accelerated. Large blur radii on big areas are the expensive ones.
No. It's displayed from memory in your browser only.
The CSS filter property applies a chain of image functions in the order written. The generator only writes a function when its value differs from the identity — brightness, contrast, saturate and opacity at 100%, grayscale, sepia and invert at 0, hue-rotate at 0deg, blur at 0 — so the output contains nothing that does no work. Percentages are written as unitless ratios (contrast(1.2) rather than contrast(120%)), which every browser accepts and which avoids rounding ambiguity; hue-rotate is in degrees and blur in pixels. drop-shadow() is offered separately because, unlike box-shadow, it follows the alpha outline of transparent PNGs and SVGs. The same chain can be emitted as backdrop-filter, with the -webkit- prefix Safari still needs, to filter whatever is behind an element instead of the element itself. Presets set several functions at once as a starting point. Your own image can be loaded for the preview; it's read into memory with an object URL and never uploaded.