GLTFLoader: metal/rough and spec/gloss materials inconsistent.

[donmccurdy]

Per KhronosGroup/glTF-Sample-Models#192, our implementation of the metal/rough and spec/gloss PBR materials is somewhat inconsistent —

In theory these two models should match. Currently most WebGL implementations do not have the spec/gloss material correct, but the BabylonJS rendering (KhronosGroup/glTF-Sample-Models#192 (comment)) looks like a good reference.

/cc @takahirox this looks similar to the roughness bug you fixed previously?

[cx20]

I confirmed that this issue has been fixed in the latest GLTFLoader.
However, the color of the background environment map seems to have changed slightly.
It used to be a color close to orange like sunset before, but seems to be whitish as a whole when using the latest version of three.js.

Is the usage wrong?

[donmccurdy]

That sounds like an effect of #15186 and #15185. If the environment map is sRGB, be sure to set texture.encoding = THREE.sRGBEncoding.

[cx20]

@donmccurdy Thank you for your advice. By setting encoding, color problem of environment map has been solved.
However, it seems that the color of the water bottle is also darkly displayed, is it correct?

[donmccurdy]

I'm not sure if that's correct or not... if other engines in the preview list are also using that environment map it would be helpful to compare.

[cx20]

I compared several engines using the same environment map. However, I did not know which was the correct display result.

[cx20]

I have created Issue on the above. #15285