Propose guidelines and approximations for real-time implementation #248
Conversation
approximation.md.html
Outdated
| Approximations for constrained implementations | ||
| ============================================== | ||
|
|
||
| This specification describes a model which includes a number of features that can be computationnally expensive and therefore impractical to support in certain contexts, like real-time applications or mobile devices. |
There was a problem hiding this comment.
Typo: Computationnally -> Computationally (double n)
There was a problem hiding this comment.
Thank you, fixed at 8416388.
approximation.md.html
Outdated
|
|
||
| The Historical background and objectives section explains how the goal of this specification is to provide an ideal, reference appearance. The implementer is, however, free to make decisions and approximations to satisfy their practical constraints. We cannot dictate what those approximations should be, as this depends on the use case. We can, however, propose some possible approximations and offer guidelines to help the implementer make those decisions. | ||
|
|
||
| As a general rule of thumb, we invite the implementer to think of those approximations in the same way one would think of different levels of detail (LOD) variants of an asset. Just like such an asset would become more coarse with distance while retaining as much detail as necessary to remain faithful to the original, an implementation of OpenPBR may decide to make simplifications, as long as the result remains reasonably faithful to the intent given the practical constraints. In the most extreme hypothetical case, the model might even get reduced to as little as a single Lambert BRDF. |
There was a problem hiding this comment.
Fixed at 8416388.
approximation.md.html
Outdated
| Layering and mixing | ||
| ------------------------------------- | ||
|
|
||
| As mentioned in the Reduction to a mixture of lobes section, the mix and layer operations can approximated as a weighted sum of BSDF lobes. |
There was a problem hiding this comment.
Typo: approximated -> approximate or be approximated
There was a problem hiding this comment.
Fixed at 8416388.
approximation.md.html
Outdated
|
|
||
| This specification describes a model which includes a number of features that can be computationally expensive and therefore impractical to support in certain contexts, like real-time applications or mobile devices. | ||
|
|
||
| The Historical background and objectives section explains how the goal of this specification is to provide an ideal, reference appearance. The implementer is, however, free to make decisions and approximations to satisfy their practical constraints. We cannot dictate what those approximations should be, as this depends on the use case. We can, however, propose some possible ones and offer guidelines to help the implementer make those decisions. |
There was a problem hiding this comment.
suggest: We don't dictate
obviously we could dictate, but choose not too
There was a problem hiding this comment.
Thank you. Reworded at fdd20b3.
Yes, most likely. I think I either didn't branch off the right point, or didn't open the PR towards the proper branch. |
virtualzavie
force-pushed
the
real-time
branch
from
fdd20b3 to
a3e56a1
Compare
I've rebased the branch on top of main, excluding the commits from the 1.x branches that were unrelated to the PR. |
|
Thanks! The diff is so much easier to look at now :) |
| Layering and mixing | ||
| ------------------------------------- | ||
|
|
||
| As mentioned in the Reduction to a mixture of lobes section, the mix and layer operations can be approximated as a weighted sum of BSDF lobes. |
There was a problem hiding this comment.
since Reduction to a mixture of lobes is not a named section in this document, can we link or otherwise direct the reader somehow?
|
|
||
| - [#McAuley15] describes a simpler and cheaper method based on [#Revie2011], consisting in modifying normals and offering a very crude approximation of anisotropic reflection. | ||
|
|
||
| - Finally, the crudest approximation would be to ignore entirely the `specular_roughness_anisotropy` for the base layer specular lobe (respectively `coat_roughness_anisotropy` for the coat specular lobe), as the anisotropy parametrization in designed to preserve the average roughness regardless of the anisotropy (see the Microfacet model section for details). |
There was a problem hiding this comment.
since the Microfacet model section is not a named section in this document, can we link or otherwise direct the reader somehow?
This PR is an attempt at proposing guidance for real-time implementation of OpenPBR, which there have been multiple requests for, including #241.
The proposal is to centralise the information relevant to real-time implementation in one place that is easy to identify and refer to in the document. There, we state clearly what the implementer is allowed to do while still complying to the specification, and we propose some approximation they can use. Ideally, we would even propose code snippets ready to use in an annex.