The type of noise to generate. Different algorithms give noise with different characteristics.

The frequency of the noise. Higher values give smaller scaled details in the noise.

The offset of the input into the noise function. If you visualize the noise as a 2D graph or 3D height field, this has the effect of “panning” across the space of possible noise outputs. If you have the general noise effect you want but just want to get a different set of values for a different look, try changing the offset.

Controls the noise’s minimum and maximum values. When you think of noise as a curve, the parameter controls the height of the peaks and valleys.

The scale increment between iterations of fractal noise added to the basic output. The higher the value the larger the “jaggies” added to the output. You can use a negative value for roughness.

Flattens the noise to prevent extreme spikes by damping the values. Higher values create a smoother look.

Controls the degree of noise. Higher values create a more chaotic and less smooth look, with more peaks and valleys.

Controls the resolution of the curl function used to generate the noise. A lower value will lead to tighter spirals in the noise.

Determines the magnitude of the effect that intervening geometry has on the created noise field. A higher value causes greater disturbances resulting from intervening geometry.

The signed distance value indicating the distance from the current position to the closest surface of intervening geometry. Piping in the result of a signed distance field with respect to a piece of geometry will allow the generated noise to react appropriately to that surface geometry.

The normal of the surface closest to the current position. Piping in a volume gradient is one way of passing in the desired normal values, and the result is a noise field that “flows” along the given surface. For best results, Distance to Surface and Surface Normal should be informed by the same surface for each position that the Curl Noise operator is performed on.

The file to load a collision field from. This will override the normal and distance values, instead, those values will be computed directly from the first volume primitive in the file. This allows the distances to vary within the noise computation, producing more accurate boundaries than if only a single sample is provided. Note the file does not have to be a disk file, it can also be an op: reference to a node or an opinput: reference to an input.

Whhen the sample point is inside the surface, the parameter will reflect the noise value in the normal direction so it is pointing outwards. This helps ensure any particles that end up inside the surface are pushed outwards by the noise field.