Image: Generators#

Noise, patterns, shapes, ramps and other image sources.

14 nodes.

Back to the reference index

Contents#

• COP Cellular Noise (PCG) — Generates cellular (Voronoi-border) noise with adjustable gaps and bevels between cells.
• COP Constant (PCG) — Creates an image filled with a single constant colour.
• COP Fill (PCG) — Fills a layer with a constant value, softly blended through an optional mask.
• COP Fractal Sample (PCG) — Layers a source image into itself at shrinking, rotating, jittered scales, like fractal noise whose basis function is an image.
• COP Gradient (PCG) — Generates a gradient (along u, v, a diagonal, radially or by angle) through a color ramp.
• COP Noise (PCG) — Generates a fractal noise image with optional tiling, animation and post-processing.
• COP Noise Extra (PCG) — Generates Bubble, Crystal, Phasor, Curl, Curl 3D and Cloud noise.
• COP Pattern (PCG) — Generates procedural patterns: checkerboard, tiles, cells, UV grid, Julia/Mandelbrot, Chladni, knit and weave.
• COP Position Map (PCG) — Generates a layer storing a 3D position per pixel, for use as the reference position of 3D noise.
• COP Random (PCG) — Generates hash-based random values per pixel or per ID.
• COP Shape (PCG) — Draws a shape as a signed distance field plus an anti-aliased mask.
• COP Stereogram (PCG) — Creates a single-image autostereogram from a depth layer.
• COP Text (PCG) — Renders text into an anti-aliased coverage mask, optionally composited over a layer.
• COP Worley Noise (PCG) — Generates Worley (cellular) noise and its feature-point, cell-border, center and ID layers.

Nodes#

COP Cellular Noise (PCG)#

PCG_COPCellularNoise · category PCG/cop · Houdini: Cellular Noise (COP), Cellular Noise 3D (COP)

Generates cellular (Voronoi-border) noise with adjustable gaps and bevels between cells.

Feature points are scattered on a lattice, each pixel finds its nearest point and the distance to the cell border. The noise value is Amplitude * shape(border) + Shift, followed by the Post Processing steps. In absolute shape mode Gap and Bevel are image-space distances; in relative mode they are fractions from 0 (cell border) to 1 (feature point). The shape is clamp((border - gap) / bevel, 0, 1): Bevel 0 gives hard flat-topped cells, larger values give pointier tops.

Besides noise the node also writes border (distance to the cell border, positive inside the cell), center (position of the nearest feature point) and id (hash of the cell), each with the Prefix prepended. If a cop is connected, its resolution is used; with a Position Layer of 2 channels you get the 2D version, with 3 channels the 3D version.

Inputs

Outputs

Parameters

Range

Pattern

Shape

Post Processing

Compared with Houdini

Covers Cellular Noise and Cellular Noise 3D. The ramp for Bevel (Bevel Ramp) is not implemented; the bevel is a linear ramp. The Houdini docs do not say whether Amplitude or Shift is applied first; here the value is multiplied first and then shifted. The Post Processing steps run in the order Fold, Complement, Bias, Gain, Gamma, Contrast, Clamp Min, Clamp Max.

COP Constant (PCG)#

PCG_COPConstant · category PCG/cop · Houdini: Constant (COP), Layer (COP)

Creates an image filled with a single constant colour.

Generates a flat colour image of the given resolution, typically as a base for compositing or as a mask backdrop. The layer has three channels (red, green, blue); enabling Use Alpha adds a fourth channel with the Alpha value. Values are not clamped, so negative and HDR values are allowed.

Outputs

Parameters

Compared with Houdini

Covers Houdini's Constant node and the constant-value use of Layer.

COP Fill (PCG)#

PCG_COPFill · category PCG/cop · Houdini: Fill (COP)

Fills a layer with a constant value, softly blended through an optional mask.

Equivalent of Fill. Every channel of the layer is replaced by the matching entry of value. If a mask layer is given, the constant is blended in proportionally to the mask (clamped to 0-1), so soft masks give soft edges; without a mask the whole layer is filled.

Inputs

Outputs

Parameters

COP Fractal Sample (PCG)#

PCG_COPFractalSample · category PCG/cop · Houdini: Fractal Sample (COP)

Layers a source image into itself at shrinking, rotating, jittered scales, like fractal noise whose basis function is an image.

Equivalent of Fractal Sample. Each octave samples the source layer with a scale of lacunarity^octave, an optional rotation (angle), jitter and random flips, around the image center. The octaves are weighted by roughness^octave and combined by the blend mode: Add, Subtract, Maximum or Over (Over only makes sense for RGBA). With one octave and no rotation or jitter the output equals the source.

Add and Subtract are normalized by the sum of the weights so that the value range matches the source. Samples first have bias subtracted (so the result is centered on 0), and offset_with_bias adds it back at the end. Over does not subtract bias.

Maintain Tileable rounds the lacunarity to an integer and the angle to multiples of 90 degrees; the source itself must already tile for the result to tile. The noise post-processing parameters (fold, complement, bias, gain, gamma, contrast, clamps) are then applied to the combined result in that order.

Inputs

Outputs

Parameters

Fractal

Randomization

Post Processing

Compared with Houdini

The weight normalization is the plugin's own choice, and the Houdini documentation's reference to gain "after weight and gain" is not implemented. Border offers clamp, wrap and black only; Filter offers bilinear and point only, with no area filtering. The optional per-pixel input layers for octaves, lacunarity and roughness are not supported.

COP Gradient (PCG)#

PCG_COPGradient · category PCG/cop · Houdini: Ramp (COP)

Generates a gradient (along u, v, a diagonal, radially or by angle) through a color ramp.

Equivalent of Ramp. The gradient position (0 to 1) is computed per pixel from the chosen direction and looked up in the ramp. The ramp uses the same text syntax as the geometry-side Attribute Fade and Measure ramps: space-separated position:value stops such as 0:0 0.5:1 1:0, or multi-component values such as 0:0,0,0 1:1,0.5,0. The number of components is taken from the first stop and sets the channel count of the layer.

Radial reaches 1 at a distance of 0.5 from the center; angular runs once around the center, from 0 to 1.

Outputs

Parameters

COP Noise (PCG)#

PCG_COPNoise · category PCG/cop · Houdini: Fractal Noise (COP), Fractal Noise 3D (COP)

Generates a fractal noise image with optional tiling, animation and post-processing.

Each pixel is given a 3D coordinate (uv times Frequency, placed on the plane z = Offset z, plus Offset x and y), and the value comes from the same fractal_noise function used by the geometry-side Attribute Noise. The same world position therefore gets a bit-identical value in an image and on geometry, as long as the parameters match, which lets you bake a displacement map and displace geometry consistently.

Channels above 1 each use their own seed (seed + k*7919), which gives RGB noise or a 2D flow field in one node (Houdini's Per Component).

Tiling: Tile U wraps the u direction onto a circle, so the result tiles horizontally. Tile U and Tile V together give a seamless 2D tile (a flat torus in four dimensions, as does the Torus noise type, which is simplex noise with both tilings). Lacunarity does not need to be an integer for tiling.

Animation: Time Scale above 0 switches to 4D noise with the fourth coordinate Time * Time Scale. Loop Length (Frames) above 0 moves time around a circle in the (z, w) plane so that frame 0 and the loop length are identical, with a flow speed matching the non-looping case. Looping needs Time Scale above 0 and cannot be combined with Tile U or Tile V, and two-way tiling cannot be combined with Time Scale either (these would need a fifth dimension; the node raises an error instead of silently disabling one).

Post processing runs after Amplitude and Center in this order: Fold, Complement, Bias, Gain, Gamma, Contrast, Clamp Minimum, Clamp Maximum. Bias, Gain and Gamma act on negative values by odd extension (sign(x) * |x|^p) so no NaN appears.

Outputs

Parameters

Range

Pattern

Pattern

Time

Fractal

Post Processing

Compared with Houdini

Covers Houdini's Fractal Noise and Fractal Noise 3D. Tile U / Tile V and Torus are implemented by wrapping the noise coordinates onto circles in a higher dimension, so the Lacunarity does not have to be rounded to an integer as in Houdini.

COP Noise Extra (PCG)#

PCG_COPNoiseExtra · category PCG/cop · Houdini: Bubble Noise (COP), Cloud Noise 3D (COP), Crystal Noise (COP), Crystal Noise 3D (COP), Curl Noise (COP), Curl Noise 3D (COP), Phasor Noise (COP)

Generates Bubble, Crystal, Phasor, Curl, Curl 3D and Cloud noise.

A collection of additional noise types. Coordinates are normalized by the image width and keep the aspect ratio, so bubbles stay round. The 3D types slice the volume at the third Offset value (z); moving it animates the noise. If cop is connected, layers are added to it and its resolution is used; otherwise a new image is created with Resolution.

The main layer is the Output Layer; auxiliary layers are named <Output Layer>_xxx:

• Bubble: one sphere per cell with probability Density and radius Radius times 0.5-1 cell; the value is the cross-section height of the sphere at the slice. Overlaps take the maximum. Auxiliary: _id.
• Crystal: 3D Voronoi where every cell has a random tilted plane, producing facets with creases between cells. Auxiliary: _id, _edge (F2-F1, zero at creases).
• Phasor: Gabor-like noise with the phase of the complex sum of kernels only, so stripe contrast is uniform. Auxiliary: _phase, _amplitude.
• Curl: a divergence-free 2-channel velocity field derived from a simplex fractal potential (unit is per noise unit; multiply by Frequency / image width for per-pixel). Auxiliary: _potential.
• Curl 3D: 3-channel curl of three potentials, evaluated on the z slice. Note that (vx, vy) on a single slice are not divergence-free.
• Cloud 3D: simplex fractal turned into a 0-1 density by Coverage and Softness. Auxiliary: _raw.

Amplitude, Center and the post-processing block only affect the main layer, not the auxiliary ones.

Inputs

Outputs

Parameters

Range

Bubble

Crystal

Phasor

Cloud

Post Processing

Compared with Houdini

Covers Houdini's Bubble, Crystal (and 3D), Phasor, Curl (and 3D) and Cloud 3D noise COPs in one node. The core algorithms of Bubble, Cloud and Crystal are the plugin's own, because Houdini does not document their formulas; Crystal's Secondary metric group and Phasor's Wave Bias are not implemented.

COP Pattern (PCG)#

PCG_COPPattern · category PCG/cop · Houdini: Cell Pattern (COP), Checkerboard (COP), Chladni Cymatic Patterns (COP), Julia Fractal (COP), Knit (COP), Labs UV Grid Texture (COP), Tile Pattern (COP), Weave (COP)

Generates procedural patterns: checkerboard, tiles, cells, UV grid, Julia/Mandelbrot, Chladni, knit and weave.

One node for eight pattern types. The main layer is the Output Layer; helper layers are named <Output Layer>_xxx and are what you use for per-element coloring and bevelled normals. If cop is connected, layers are added to it and its resolution is used; otherwise a new image of the given Resolution is created. When the counts are whole numbers the patterns tile seamlessly.

Pattern types and their layers:

• Checkerboard: analytically antialiased 0/1; _id.
• Tile Pattern: bricks or tiles with a Gap and Bevel given as fractions of a tile's short side. Layers: main mask (gaps are 0), _id, _height (bevel), _edge (pixel distance to the gap), _rand (per tile), _uv (0-1 within the tile). Bond sets the brick layout; seamless tiling needs an even row count (American: a common multiple of Header Interval and 2) and a column count that is a whole multiple of the bond's row period.
• Cell Pattern: periodic Voronoi cells. Layers: main mask, _id, _edge (true pixel distance to the cell border, measured in pixel space), _dist (distance to the feature point, normalized by the short cell side), _rand.
• UV Grid: a UV checking image; cells are named from A1 at the bottom-left of UV space (v up), so a glance at the textured model shows mirroring or rotation. Layers: RGB main, _id, _uv.
• Julia Fractal: Julia or Mandelbrot with smooth iteration count; view width is 3/Zoom complex units. Layers: main (smooth iterations divided by Max Iterations), _inside, _trap (orbit trap).
• Chladni Cymatics: the standing-wave pattern of a square plate, f = A*[cos(n*pi*x)cos(m*pi*y) +/- k*cos(m*pi*x)cos(n*pi*y)], where A is Amplitude and k is Amplitude Ratio. Sand collects on the nodal lines (f = 0). Layers: main, _value (signed shape).
• Knit: stockinette V-shaped loops. Layers: height, _mask, _id, _leg (0 left, 1 right), _rand.
• Weave: warp (vertical) and weft (horizontal) threads with half-round cross-section that undulate over and under at crossings. Layers: height, _mask, _warp (where the warp is on top), _id, _rand.

Inputs

Outputs

Parameters

Tile

Cell

Weave

Julia

Chladni

UV Grid

Compared with Houdini

Covers Houdini's Checkerboard, Tile Pattern, Cell Pattern, Labs UV Grid Texture, Julia Fractal, Chladni Cymatic Patterns, Knit and Weave. Knit's geometry is the plugin's own construction.

COP Position Map (PCG)#

PCG_COPPositionMap · category PCG/cop · Houdini: Position Map (COP)

Generates a layer storing a 3D position per pixel, for use as the reference position of 3D noise.

Each axis is computed as (P + Shift) * Cycles and then folded back into the [-1, 1] range according to that axis' Border mode.

There is no camera: Position is (image x, image y, 0) and View Direction is (0, 0, -1). Without a connected reference image the Resolution parameter sets the size.

Inputs

Outputs

Parameters

X

Y

Z

Compared with Houdini

There is no camera, so Camera Origin is not implemented.

COP Random (PCG)#

PCG_COPRandom · category PCG/cop · Houdini: Random Mono (COP), Random RGB (COP)

Generates hash-based random values per pixel or per ID.

Per pixel gives white noise; per ID gives the same value to every pixel sharing an ID, which colours each region of a segmentation image randomly. The hash makes results reproducible bit for bit. Without a connected image a new one is created at the Resolution (per-pixel only).

Shared Min/Max uses one range for all channels. Min/Max, Color Ramp and Specific Values are the three modes of Houdini's Random RGB and always output RGB (with Channels 4 the alpha is constantly 1).

Inputs

Outputs

Parameters

Range

Random

COP Shape (PCG)#

PCG_COPShape · category PCG/cop · Houdini: SDF Shape (COP)

Draws a shape as a signed distance field plus an anti-aliased mask.

The primary output is an SDF (negative inside, positive outside), not a black and white image; the mask is derived from it with a smoothstep. This matters: an SDF directly supports outlines (|d| < w), growing and shrinking (d - r) and bevels; anti-aliasing is analytic, with the edge width given in pixels so softness looks the same at any resolution; and min / max of two SDFs gives union / intersection while staying a valid distance field.

Coordinates are centred and normalized by the short edge: the short side spans [-0.5, 0.5] and the long side extends by the aspect ratio. A circle of radius 0.5 is therefore inscribed in the short side at any aspect, and never squashed into an ellipse.

Rectangles use the standard exact SDF. The star is not an exact SDF (a polar fold gives an approximation), which is fine for outlines and masks but visible with large offsets.

The layers are only produced if their names are not empty. If Color Layer is set, a 3-channel copy of the mask is written to it.

Outputs

Parameters

Circle

Box

Star

Shape

Line

Compared with Houdini

Based on Houdini's SDF Shape, with the Shape Class level removed: all shapes are in one menu. Spiral, Wave, Fishscale, Octagon-dot, Horseshoe, Quadratic Circle and the Marker shapes are not implemented.

COP Stereogram (PCG)#

PCG_COPStereogram · category PCG/cop · Houdini: Auto Stereogram (COP)

Creates a single-image autostereogram from a depth layer.

Turns depth (0 far, 1 near) into a stereogram made of random dots or of a repeated pattern layer tiled at the strip width. The default is for parallel viewing; Cross-Eyed reverses it. Hidden surface removal is not done, so ghosting can appear at abrupt depth edges. A per-pixel separation layer in pixels is also written as <Output Layer>_sep.

Inputs

Outputs

Parameters

COP Text (PCG)#

PCG_COPText · category PCG/cop · Houdini: Font (COP)

Renders text into an anti-aliased coverage mask, optionally composited over a layer.

Draws Text with FreeType anti-aliasing into Output Layer as a 0-1 coverage mask. Position is a UV point (V pointing up) and Align / VAlign choose which point of the text block is placed on it. Multi-line text is supported, with line distance set by Line Spacing.

If Composite Layer is set, Color is laid over that layer through the mask (a black layer is created if it does not exist yet). The font size is in pixels, so it does not follow resolution changes automatically.

The empty font uses the bundled Pillow font, which only has Latin characters; for other scripts give the file name of a system font that covers them.

Inputs

Outputs

Parameters

COP Worley Noise (PCG)#

PCG_COPWorleyNoise · category PCG/cop · Houdini: Worley Noise (COP), Worley Noise 3D (COP)

Generates Worley (cellular) noise and its feature-point, cell-border, center and ID layers.

One feature point is placed in each cell of a lattice (jittered by Jitter) and every pixel measures distances to the nearest points. Five layers are written, named with Prefix plus:

• dist1: distance to the nearest feature point.
• dist2: distance to the second nearest point.
• border: distance to the nearest cell border, positive inside the cell. Exact for the Euclidean metric; for Manhattan and Chebyshev it is the estimate (dist2 - dist1) / 2.
• center: coordinates of the nearest feature point.
• id: hash of the cell holding the nearest point, as a signed integer.

Coordinates are normalised by image width (x = u, y = v x height/width). If an input is connected and Pos Layer is set, its values are used as coordinates: 2 channels give the 2D version, 3 channels the 3D version (Worley Noise 3D), for example world positions. The post-processing section (fold, complement, bias, gain, gamma, contrast, clamp) is applied only to dist1 and dist2.

Distances are in image coordinates (times Element Size); with an anisotropic Element Scale they are measured before the stretch. The search covers a 5x5 (2D) or 5x5x5 (3D) neighbourhood of cells, which is sufficient for Jitter up to 1; larger jitter can give artefacts, as in Houdini.

Inputs

Outputs

Parameters

Pattern

Post Processing

Compared with Houdini

Covers Worley Noise and Worley Noise 3D. Image coordinates are normalised by the image width. Houdini states that border is exact for all metrics; here it is exact only for Euclidean.

Tips

• Use border as a mask for cracks, or id to give every cell a random value (hash it in a Wrangle).