🍄Great Random Organic Gradient

The name tells you almost everything: give it a size, a few colors, and a blob count, and it hands you a soft, painterly gradient image plus the exact color palette it used. If you've ever wasted ten minutes hunting for a background that isn't the same flat gray everyone else uses, this is the node for you.

It's one of five nodes in the small NDDG_Great_Nodes pack (everything is prefixed with a mushroom emoji, the author clearly likes them). This one is the "Generators" entry: a procedural background factory. The wider point of the pack is that its pieces chain together - you can generate an organic gradient here, then blend it over a render with the pack's ImageBlend node. Standalone it's also handy as a color reference generator, because it doesn't just give you an image.

How it works

Behind the scenes it's exactly what you'd hope: a seeded numpy random number generator picks blob positions, sizes, and colors, draws circles, squares, random polygons, or soft radial spots onto a canvas, and then runs a Gaussian blur over the whole thing until the shapes melt into an organic wash. The blur_strength is the main dial for how "organic" it feels - low values give you visible blobs, high values give you an almost airbrushed field of color.

The part people don't expect is the trio of outputs. Besides the image itself, you get a palette_image (a horizontal swatch strip showing every color in play) and palette_hex (those colors as a comma-separated string like #FF3300, #00FFE1). That string is surprisingly useful: paste it into a prompt, feed it to a color-conditioning node, or just keep it as a record of what you used. The image comes out with an alpha channel, so it layers cleanly over other stuff.

The inputs that matter

You'll set most of these once and forget them, but three are worth understanding:

• seed - default -1 means "randomize every run," which is fun until you find a gradient you love and can't reproduce it. Set a real number to lock it in. Change the seed for a new variation.
• colors (2–8) and blob_count (1–50) - the palette size and how many blobs get drawn. More blobs plus more blur equals a denser, more textured background.
• random_palette - on by default, which means it picks the colors for you. Flip it off and the color1…color8 hex fields become your manual palette (only as many as colors asks for get used).

blob_shape gives you circle, square, polygon, or radial; transparent_background and random_background control what's behind the blobs. There's a color-picker widget on the manual color fields courtesy of the pack's bundled JavaScript, so you don't have to type hex codes blind.

How to install it

Install the whole pack - all five nodes ship together in NDDG_Great_Nodes:

• ComfyUI Manager → "Install Custom Nodes" → search for NDDG_Great_Nodes (or "Great Nodes"), install, restart.
• Or the manual route:

cd ComfyUI/custom_nodes
git clone https://github.com/NeoDroleDeGueule/NDDG_Great_Nodes

then restart ComfyUI. There's no requirements.txt and no model downloads - this is a pure Python node using the PIL/numpy/torch that ship with ComfyUI, so it's one of the friendlier installs in the ecosystem. The one caveat is that the pack imports matplotlib at load time for its sigmas preview, so if your ComfyUI Python env lacks it (some bare venv setups do), you'll see an import error and nothing from the pack will appear. Fix it with pip install matplotlib.

Common issues

There aren't many failure modes here because it's a simple generator, but the two things that bite people:

• "Why is it different every run?" - because seed is -1 by default. Set a fixed seed, or note the palette_hex output if you need to rebuild a specific look.
• "My colors are ignored!" - random_palette defaults to true. Turn it off before you start tweaking color1 through color8, or it'll cheerfully overwrite your choices.

It's not a node that will change your life - but it's the kind of small, reliable utility that makes ComfyUI feel less like a chore. Generate a gradient, grab the palette hex, and go.