In RenderMan, in addition to the built-in AOVs, light path expressions (LPEs), which are adopted from Open Shading Language's Light Path Expression, specify what light transport paths to output to a display channel. In other words, we don't have to modify the shaders or plugins to output a custom AOV that uses the LPE.
The setup is the same as an AOV except we use a light path expression instead of an AOV channel name. Additionally, we can group the outputs by light groups and/or by geometry set up by a lpegroup.
You will find a useful set of video tutorials and discussion of these practices on the RenderMan Community site.
To better understand how an LPE is constructed, the below tokens are used to define interactions.
<type event 'group'>
|Scattering event specified by scattering type, scattering event, or lpe group|
|Light, i.e. analytic and geometric lights|
|Emissive object, i.e. camera visible lights and material glow|
|Diffuse (includes D1 ... D4)|
|Specular (includes S1 ... S8)|
|User signal (includes U1 ... U12)|
|( )||Parentheses allow grouping of a sequence of events|
||||A vertical bar separates alternatives. For example |
|[ ]||Square brackets will match any one of the enclosed events. For example |
|Match preceding event zero or more times. For example |
|Match one or more times. For example |
|Match exactly n times|
|Match min or more times|
|Match at least min, but not more than max times.|
To specify an LPE in RIB, reference a DisplayChannel in the Display. This is the most basic way to view what's being output.
LPE tokens are not necessarily the most artist-friendly things. For this reason, we provide a list of LPEs in our plugins that can be easily selected and output. We also provide options on how specific you want to be when it comes to output; from outputting all your direct or indirect interactions at once per type or outputting them per-lobe which provides greater control for advanced workflows.
For basic workflows, the following LPE are recommended. Note that we are specifying the DisplayChannel and a name (which can be most anything). Note that these LPE will collect the light interaction from different lobes of materials and store them in the same AOV. For example: reflections from the Glass Lobe in PxrSurface will be combined with other specular reflections in the direct and indirectSpecular AOVs. This can simplify the outputs.
Understand that emissive LPE results are a special case. In the materials provided it's referred to as "Glow" or emission. Light sources directly visible to the camera are also output into this LPE. In the below example, the candle flames are using the glow lobe in PxrSurface.
RenderMan Still Life by Dylan Sisson
Bxdf materials may have
S, respectively. However, in some situations, it may be desirable to output a specific lobe separately.
For example, some PxrSurface usage may involve the Clearcoat lobe. Normally this is summed under the
S token. Routing this lobe to the
S2 token will allow you to use
S2 in your LPEs. Up to 4 diffuse lobes and 8 specular lobes are available for LPEs. By default,
S1 contains all the specular lobes, so use any higher tokens such as
S2 for your per-lobe LPE's.
For advanced workflows that want to separate all scattering lobes into separate AOVs, the below LPE are provided. These capture the light interaction for each lobe, direct and indirect, and store them into separate AOVs. This is useful for tweaking individual effects at the cost of added AOVs and image output.
Note that the emissive LPE is repeated below. For materials in RenderMan it's typically called Glow.
Caustics are not included in the built-in list because they overlap with
indirectdiffuse lpe:C<RD>[DS]+[<L.>O] However, you can select caustics with the expression
lpe:CDS+[<L.>O] Below you can find an example of the caustics LPE output, Indirect Diffuse, and the difference between them. Note that the Caustics LPE output doesn't capture indirect lighting from object to object.
If a custom LPE is useful for your shots, you can add your custom LPE by creating a macro in
rendermn.ini file. The custom LPE will appear in the RenderMan for Maya (RfM) LPE list after relaunching Maya. For example:
- It is important to use
<L.>instead of only
Lso that it will be compatible with light groups. Otherwise, the renderer will output a warning.
lpe:prefix will be automatically added to your macro's name when the file is parsed.
An LPE group (lpegroup) allows us to specify which objects we want to be use for an LPE channel. The attributes can be added in RenderMan plugins using the software's menu system such as PrmanObjectSettings in Katana.
LPE Group Example RIB
Predefined LPE Groups
For the built-in LPE such as shadowcollector, it assumes a predefined lpegroup named "collector". So for the objects that we want to collect shadow, we can simply name its lpegroup to "collector". Specifying
lpe:shadowcollector will collect the shadow for these objects.
This may be useful for creating holdouts using a manual workflow.
Advanced LPE Group Logic
You can use the usual regular expression character classes syntax to define more complex LPEs by considering each lpegroup as a single token.
For instance, provided you are using lpegroups "foo" and "bar", you can define the following LPEs:
By default, LPEs retrieve the response to all lights in the scene. To limit the LPE to the contribution due to a single light or set of lights, assign each light to the group by setting the __group parameter on each light to the same name.
For the short LPE name, suffix the name with '_' followed by the light group name. For the long expression, place the light group name in single quotes inside the LPE.
For instance, say we have a light group named "key". The following are both valid LPEs:
unoccluded– returns unoccluded or unshadowed result.
noclamp– returns unclamped result.
nothruput– does not apply thruput (thruput is the accumulative albedo of the objects hit by rays).
shadows– returns collected shadows.
holdout– holdout LPE that directs the LPE output to the beauty channel.
overwrite– instead of outputting the accumulated result, overwrite it. One example of using this is for the albedo output where we do not want an accumulated result.
noinfinitecheck– do not do any infinite check.
User Defined Signals
Utokens specify a user defined signal. A Bxdf can output any user defined signal via the RixBXLobeWeights class. There are twelve of them, from
U12. By default, all user defined signals are set to
U1. More on this can be found in the Developer's Guide.
- If you have more than one user defined signal, it is important to direct each user defined signal to a different
Utoken by setting them in
rendermn.ini. Otherwise, all user defined signals that are assigned to the same U token will override each others!
- By default,
U2is set to the albedo output in the installed
- <L.> means Light scattering type with all scattering events. For light group AOVs, using <L.> is required instead of L.
- lpe:CD<L.> means camera with diffuse scattering event for light scattering type.
- lpe:CO means camera with emissive objection (since emissive does not need any light, there is no L).