# MaterialX: Proposed Additions and Changes **Proposals for Version 1.39** September 15, 2024 # Introduction The [MaterialX Specification](./MaterialX.Specification.md) has historically included descriptions of not just current functionality, but also forward-looking proposed functionality intended for eventual implementation. We believe it will be beneficial to provide clarity on which functionality is currently supported in the library, and which sections document proposed additions. As such, those forward-looking proposals have been moved from the formal Specification documents into this Proposed Additions and Changes document to be discussed and debated. These descriptions can then be migrated into the appropriate formal Specification document once actually implemented in the code base. New proposals for changes and additions to MaterialX may be added to this document once a generally favorable consensus from the community is reached. ## Table of Contents **[Introduction](#introduction)** **[Proposals: General](#propose-general)** **[Proposals: Elements](#propose-elements)** **[Proposals: Stdlib Nodes](#propose-stdlib-nodes)** **[Proposals: PBR Nodes](#propose-pbr-nodes)** **[Proposals: NPR Nodes](#propose-npr-nodes)**
# Proposals: General ## Color Spaces When the OCIO NanoColor library (provide link) becomes available, MaterialX should support the official colorspace names in that spec, with the current MaterialX colorspace names supported as aliases. MaterialX should also support the following color spaces: * `lin_rec2020` * `g22_rec2020`
# Proposals: Elements
### AOV Output Elements
A functional nodegraph with either a "shader" or "material"-semantic output type may contain a number of <aovoutput> elements to declare arbitrary output variables ("AOVs") which the renderer can see and output as additional streams of information. AOVoutputs must be of type float, color3 or vector3 for pre-shading "pattern" values, or BSDF or EDF for shader-node output values; the renderer is expected to extract the appropriate color-like information from BSDF and EDF types. AOVs defined within a shader-semantic node instantiated within this functional nodegraph may be "passed along" and potentially renamed (but may not be modified or operated on in any way) by providing a sourceaov attribute in the <aovoutput>.
```xml
# Proposals: Stdlib Nodes
### Procedural Nodes
* **`tokenvalue`**: a constant "interface token" value, may only be connected to <token>s in nodes, not to <input>s.
* `value` (any uniform non-shader-semantic type): the token value to output; "enum" and "enumvalues" attributes may be provided to define a specific set of allowed token values.
### Noise Nodes
We have a standard 3d fractal noise, but a 2d variant would be useful as well.
* **`fractal2d`**: Zero-centered 2D Fractal noise in 1, 2, 3 or 4 channels, created by summing several octaves of 2D Perlin noise, increasing the frequency and decreasing the amplitude at each octave.
* `amplitude` (float or vectorN): the center-to-peak amplitude of the noise (peak-to-peak amplitude is 2x this value). Default is 1.0.
* `octaves` (integer): the number of octaves of noise to be summed. Default is 3.
* `lacunarity` (float or vectorN): the exponential scale between successive octaves of noise; must be an integer value if period is non-zero so the result is properly tileable. VectorN-output types can provide either a float (isotropic) or vectorN (anisotropic) values for lacunarity and diminish. Default is 2.0.
* `diminish` (float or vectorN): the rate at which noise amplitude is diminished for each octave. Should be between 0.0 and 1.0; default is 0.5. VectorN-output types can provide either a float (isotropic) or vectorN (anisotropic) values for lacunarity and diminish.
* `period` (float or vectorN): the positive integer distance at which the noise function returns the same value for texture coordinates repeated at that step. Default is 0, meaning the noise is not periodic.
* `texcoord` (vector2): the 2D texture coordinate at which the noise is evaluated. Default is to use the first set of texture coordinates.
1D Cell noise was proposed an an alternative approach to random value generation.
* **`cellnoise1d`**: 1D cellular noise, 1 or 3 channels (type float or vector3).
* `period` (float or vector3): the positive integer distance at which the noise function returns the same value for input coordinate repeated at that step. Default is 0, meaning the noise is not periodic.
* `in` (float): the 1D coordinate at which the noise is evaluated.
Expanded 2D Worley noise to support different distance metrics and periodicity.
* **`worleynoise2d`**: 2D Worley noise using centered jitter, outputting float (distance metric to closest feature), vector2 (distance metrics to closest 2 features) or vector3 (distance metrics to closest 3 features).
* `metric` (uniform string): the distance metric to return, one of "distance" (Euclidean distance to feature), "distance2" (Euclidean distance squared), "manhattan" or "chebyshev". Default is "distance".
* `period` (float or vector3): the positive integer distance at which the noise function returns the same value for texture coordinates repeated at that step. Default is 0, meaning the noise is not periodic.
Expanded 3D Worley noise to support different distance metrics and periodicity.
* **`worleynoise3d`**: 3D Worley noise using centered jitter, outputting float (distance metric to closest feature), vector2 (distance metrics to closest 2 features) or vector3 (distance metrics to closest 3 features).
* `metric` (uniform string): the distance metric to return, one of "distance" (Euclidean distance to feature), "distance2" (Euclidean distance squared), "manhattan" or "chebyshev". Default is "distance".
* `period` (float or vector3): the positive integer distance at which the noise function returns the same value for position coordinates repeated at that step. Default is 0, meaning the noise is not periodic.
#### Periodic Noises
In #1201 it was decided that separate periodic versions of all of the noises is preferred to adding it to the existing noises.
### Shape Nodes
### Geometric Nodes
* **`bump`**: Existing node, proposal to add a vector3 `bitangent` input
Note: when <geompropvalueuniform> is added, the text in the first paragraph of the Specification about Node Inputs should be revised to include "<geompropvalueuniform>" as an example of "or any other node whose output is explicitly declared to be uniform".
### Global Nodes
* **`ambientocclusion`**: Compute the ambient occlusion at the current surface point, returning a scalar value between 0 and 1. Ambient occlusion represents the accessibility of each surface point to ambient lighting, with larger values representing greater accessibility to light. This node must be of type float.
* `coneangle` (float): the half-angle of a cone about the surface normal, within which geometric surface features are considered as potential occluders. The unit for this input is degrees, and its default value is 90.0 (full hemisphere).
* `maxdistance` (float): the maximum distance from the surface point at which geometric surface features are considered as potential occluders. Defaults to 1e38, e.g. "unlimited".
### Application Nodes
* **`updirection`**: the current scene "up vector" direction, as defined by the shading environment. This node must be of type vector3.
* `space` (uniform string): the space in which to return the up vector direction, defaults to "world".
### Math Nodes
* **`transformcolor`**: transform the incoming color from one specified colorspace to another, ignoring any colorspace declarations that may have been provided upstream. For color4 types, the alpha channel value is unaffected.
* `in` (color3 or color4): the input color.
* `fromspace` (uniform string): the name of a standard colorspace or a colorspace understood by the application to transform the `in` color from; may be empty (the default) to specify the document's working colorspace.
* `tospace` (uniform string): the name of a standard colorspace or a colorspace understood by the application to transform the `in` color to; may be empty (the default) to specify the document's working colorspace.
* **`triplanarblend`** (NG): samples data from three inputs, and projects a tiled representation of the images along each of the three respective coordinate axes, computing a weighted blend of the three samples using the geometric normal.
* `inx` (float or colorN): the image to be projected in the direction from the +X axis back toward the origin. Default is 0 in all channels.
* `iny` (float or colorN): the image to be projected in the direction from the +Y axis back toward the origin with the +X axis to the right. Default is 0 in all channels.
* `inz` (float or colorN): the image to be projected in the direction from the +Z axis back toward the origin. Default is 0 in all channels.
* `position` (vector3): a spatially-varying input specifying the 3D position at which the projection is evaluated. Default is to use the current 3D object-space coordinate.
* `normal` (vector3): a spatially-varying input specifying the 3D normal vector used for blending. Default is to use the current object-space surface normal.
* `blend` (float): a 0-1 weighting factor for blending the three axis samples using the geometric normal, with higher values giving softer blending. Default is 1.0.
* `filtertype` (uniform string): the type of texture filtering to use; standard values include "closest" (nearest-neighbor single-sample), "linear", and "cubic". If not specified, an application may use its own default texture filtering method.
### Adjustment Nodes
* **`curveinversecubic`**: remap a 0-1 input float value using an inverse Catmull-Rom spline lookup on the input `knots` values. Outputs a 0-1 float interpolant value.
* `in` (float): the input value or nodename
* `knots` (uniform floatarray): the list of non-uniformly distributed input values defining the curve for the remapping. At least 2 values must be provided, and the first and last knot have multiplicity 2.
* **`curveuniformlinear`**: output a float, colorN or vectorN value linearly interpolated between a number of `knotvalues` values, using the value of `in` as the interpolant.
* `in` (float): the input interpolant value or nodename
* `knotvalues` (uniform floatarray or colorNarray or vectorNarray): the array of at least 2 values to interpolate between.
* **`curveuniformcubic`**: output a float, colorN or vectorN value smoothly interpolated between a number of `knotvalues` values using a Catmull-Rom spline with the value of `in` as the interpolant.
* `in` (float): the input interpolant value or nodename
* `knotvalues` (uniform floatarray or colorNarray or vectorNarray): the array of at least 2 values to interpolate between.
* **`curveadjust`** (NG): output a smooth remapping of input values using the centripetal Catmull-Rom cubic spline curve defined by specified knot values, using an inverse spline lookup on input knot values and a forward spline through output knot values. All channels of the input will be remapped using the same curve.
* `in` (float or colorN or vectorN): the input value or nodename
* `numknots` (uniform integer): the number of values in the knots and knotvalues arrays
* `knots` (uniform floatarray): the list of input values defining the curve for the remapping. At least 2 and at most 16 values must be provided.
* `knotvalues` (uniform floatarray): the list of output values defining the curve for the remapping. Must be the same length as knots.
* **`curvelookup`** (NG): output a float, colorN or vectorN value smoothly interpolated between a number of knotvalue values, using the position of in within knots as the knotvalues interpolant.
* `in` (float): the input interpolant value or nodename
* `numknots` (uniform integer): the number of values in the knots and knotvalues arrays
* `knots` (uniform floatarray): the list of knot values to interpolate in within. At least 2 and at most 16 values must be provided.
* `knotvalues` (uniform floatarray or colorNarray or vectorNarray): the values at each knot position to interpolate between. Must be the same length as knots.
### Compositing Nodes
### Conditional Nodes
* **`ifelse`**: output the value of one of two input streams, according to whether the value of a boolean selector input is true or false
* `infalse`, `intrue` (float or colorN or vectorN): the values or nodenames to select from based on the value of the `which` input. The types of the various `inN` inputs must match the type of the `switch` node itself. The default value of all `inN` inputs is 0.0 in all channels.
* `which` (boolean): a selector to choose which input to take values from; default is "false".
### Channel Nodes
* **`extractrowvector`**: extract the specified row vector number from a matrixN stream.
* `in` (matrixN): the input value or nodename
* `index` (integer): the row number to extract, should be 0-2 for matrix33 streams, or 0-3 for matrix44 streams.
* **`separatecolor4`** (NG): output the RGB and alpha channels of a color4 as separate outputs.
* `in` (color4): the input value or nodename
* `outcolor` (output, color3): the RGB channel values.
* `outa` (output, float): the value of the alpha channel.