Closed pushmatrix closed 5 years ago
Thanks for reporting this @pushmatrix ! We'll take a look
FYI: I updated the example given that shows the difference even more.
Okay, so looking at your metallic-roughness map, it looks like you're going for metal = 0 uniformly and roughness a bit noisy, between about 130-155? Inspecting threejs I'm seeing the metalnessMap and roughness map getting the same texture, and maybe some other issues. We might need to check out our gltf loader. I notice you have some indirection here, where scene.gltf encapsulates fabricball.glb; does this still repro if everything is baked into a single gltf?
Yep that's what we're going for. Same thing happens with it baked into one. Here's the gltf
Not sure how the indirection happened. It was originally a .glb from substance then I exported the .gltf + textures from Sketchfab after uploading it.
But yes you're right it does look like it has metalness applied to it.
Playing around with setting the metalness map in Sketchfab to be the same as the roughness, and the model didn't look the same as model viewer:
However setting the roughness multiplier to 0.5 and then it looked pretty similar to model-viewer:
I take it back; metalness looks fine. I'm now investigating what I think is a difference between roughness and roughness squared.
Filament does seem a bit less shiny:
@elalish Is the rendering issue here solved by #591 or did #591 just make it so PMREM is on by default? When this issue was first posted I had mistakenly not turned on PMREM, but then updated the issue to show the fabric ball being way too shiny even with PMREM on.
Does #591 make it less shiny?
No, but I’m also working on that. That’s a bigger change affecting the internals of three js materials and shaders.
On Mon, Jul 22, 2019 at 7:09 PM Daniel Beauchamp notifications@github.com wrote:
@elalish https://github.com/elalish Is the rendering issue here solved by #591 https://github.com/GoogleWebComponents/model-viewer/pull/591 or did #591 https://github.com/GoogleWebComponents/model-viewer/pull/591 just make it so PMREM is on by default? When this issue was first posted I had mistakenly not turned on PMREM, but then updated the issue to show the fabric ball being way too shiny even with PMREM on.
Does #591 https://github.com/GoogleWebComponents/model-viewer/pull/591 make it less shiny?
— You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub https://github.com/GoogleWebComponents/model-viewer/issues/587?email_source=notifications&email_token=AAMS2LHGDL2BEPGQUFN67A3QAZR5RA5CNFSM4HR33KG2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD2RWHJA#issuecomment-514024356, or mute the thread https://github.com/notifications/unsubscribe-auth/AAMS2LDIZ62T6WL7RQM2TSTQAZR5RANCNFSM4HR33KGQ .
Is it ok if we reopen this issue then?
Sorry about letting this get caught in the issue-closing crossfire @pushmatrix
Description
I'm having an issue with the lighting on certain PBR models that have a lot of small details in the normal map. The way the details in the normal map are reflecting light seems to be off and the model comes across a bit too shiny and wet.
Below are images in various viewers. They use the same environment map from here: https://github.com/GoogleWebComponents/model-viewer/issues/521#issuecomment-488694339.
Model Viewer
Sketchfab
Substance
AR Quick Look
(Doesn't use the same environment map)
We created the material in substance painter, and Sketchfab & AR Quick Look render it as designed. Model-viewer makes it look like the normal map roughness is really low and reflective.
Live Demo
https://ninth-paste.glitch.me/
Download the model here
Browser Affected
OS
Versions