MatCap Improvement #104640

New Issue

Clément Foucault · 2023-02-12T00:52:52+01:00

Clément Foucault commented

2023-02-12 00:52:52 +01:00

Separate Specular Layer

The matcaps included in blender were introduce in the 2.8 version. At this time, there was concept of specular layer.
A patch was submitted (need to find it) to split the specular layer from the diffuse layer for existing internal matcaps. Unfortunately the patch was corrupted and we need to do the work all over again. But we can go one step further.

Color Neutrality

For correct support of all material parameters, the diffuse and specular components need to be color neutral. This means the surface color should be a white surface. It doesn't mean the whole matcap needs to be grayscale. Colored light can still bounce off the white surface. The surface can also have some kind of subsurface scattering.

Metallic support

For good support of metallic surfaces, there needs to be some variety to reflect in the specular channel. However, to not interfere with older matcaps behavior, we might consider making a different EXR layer only for metal reflection support.

It might prove complex to handle matcaps with metal reflection in the diffuse channel, not to mention if they have a clear-coat reflection. A possible solution would be to only color the specular reflexion (no scalling) and mix down the diffuse a bit by the amount of energy there is in the specular channel.

Roughness Support

Roughness support can be added by pre-filtering the specular layer and store that in the mipmap chain. This can be done automatically by workbench.
Ideally recorded roughness should be 0.0 with so that default roughness 0.5 is soft enough. However, this feels like highly constraining the freedom of matcap authoring.

Maybe an ad-hoc system could be figured out to "sharpen" the reflections (to a small extent, we don't want to loose the matcap appearance). This would recover the 0.0 to 0.5 roughness range.

Another option is to remap the roughness to a non-linear value that is almost starting at 0.5 (see attachment). This way, the default matcap appearance would be very close to what the matcap looks. However, this makes the roughness parameter feel weird.

Automatic conversion for other matcaps

For user matcaps, we could come up with an auto setup for auto splitting the diffuse and specular light. This would improve all other color modes other than the default white.
However this separation might be difficult for good looking metals. In the event where no good enough solution is found, we can disable the metallic parameter when using these matcaps. Parameters could be added to the studio-lights to help get a more proper split.

Main light extraction

This would be needed for the shadow to follow the light direction #xxxx as well as lighting the volumes #xxxx.
A Spherical Harmonic L1 extraction should provide an ambient and a directional term.

### Separate Specular Layer The matcaps included in blender were introduce in the 2.8 version. At this time, there was concept of specular layer. A patch was submitted (need to find it) to split the specular layer from the diffuse layer for existing internal matcaps. Unfortunately the patch was corrupted and we need to do the work all over again. But we can go one step further. ### Color Neutrality For correct support of all material parameters, the diffuse and specular components need to be color neutral. This means the surface color should be a white surface. It doesn't mean the whole matcap needs to be grayscale. Colored light can still bounce off the white surface. The surface can also have some kind of subsurface scattering. ### Metallic support For good support of metallic surfaces, there needs to be some variety to reflect in the specular channel. However, to not interfere with older matcaps behavior, we might consider making a different EXR layer only for metal reflection support. It might prove complex to handle matcaps with metal reflection in the diffuse channel, not to mention if they have a clear-coat reflection. A possible solution would be to only color the specular reflexion (no scalling) and mix down the diffuse a bit by the amount of energy there is in the specular channel. ### Roughness Support Roughness support can be added by pre-filtering the specular layer and store that in the mipmap chain. This can be done automatically by workbench. Ideally recorded roughness should be 0.0 with so that default roughness 0.5 is soft enough. However, this feels like highly constraining the freedom of matcap authoring. Maybe an ad-hoc system could be figured out to "sharpen" the reflections (to a small extent, we don't want to loose the matcap appearance). This would recover the 0.0 to 0.5 roughness range. Another option is to remap the roughness to a non-linear value that is almost starting at 0.5 (see attachment). This way, the default matcap appearance would be very close to what the matcap looks. However, this makes the roughness parameter feel weird. ### Automatic conversion for other matcaps For user matcaps, we could come up with an auto setup for auto splitting the diffuse and specular light. This would improve all other color modes other than the default white. However this separation might be difficult for good looking metals. In the event where no good enough solution is found, we can disable the metallic parameter when using these matcaps. Parameters could be added to the studio-lights to help get a more proper split. ### Main light extraction This would be needed for the shadow to follow the light direction #xxxx as well as lighting the volumes #xxxx. A Spherical Harmonic L1 extraction should provide an ambient and a directional term.

Capture d’écran 2023-02-13 à 15.15.57.png

199 KiB

❤️ 1

Clément Foucault added the

Type

To Do

label 2023-02-12 00:52:52 +01:00

Clément Foucault added this to the EEVEE & Viewport project 2023-02-12 00:52:53 +01:00

TheRedWaxPolice commented

2023-02-12 02:15:46 +01:00

Hi... Any hope on using different matcaps on different objects?

Clément Foucault commented

2023-02-13 01:47:23 +01:00

That's not what this task is about. That would be a workflow change and this task is only about cleaning up / improving existing matcap rendering.

Note that the different matcaps approach is what we started with during early 2.8 development and we reverted back to single matcap for the whole scene. I don't remember the reason why.

That's not what this task is about. That would be a workflow change and this task is only about cleaning up / improving existing matcap rendering. Note that the different matcaps approach is what we started with during early 2.8 development and we reverted back to single matcap for the whole scene. I don't remember the reason why.

TheRedWaxPolice commented

2023-02-13 03:20:18 +01:00

That's not what this task is about. That would be a workflow change and this task is only about cleaning up / improving existing matcap rendering.

Note that the different matcaps approach is what we started with during early 2.8 development and we reverted back to single matcap for the whole scene. I don't remember the reason why.

Alright, no problem..
When I read the title "MatCap Improvement" I thought yay this gotta be it haha...
I hope it lands someday... to me that's the biggest downside of using matcaps in blender... 👍

> That's not what this task is about. That would be a workflow change and this task is only about cleaning up / improving existing matcap rendering. > > Note that the different matcaps approach is what we started with during early 2.8 development and we reverted back to single matcap for the whole scene. I don't remember the reason why. Alright, no problem.. When I read the title "MatCap Improvement" I thought *yay* this gotta be it haha... I hope it lands someday... to me that's the biggest downside of using matcaps in blender... 👍

Julien Kaspar added the

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

MatCap Improvement #104640