OpenPBRMaterial (including loading and exporting glTF) #16773

MiiBond · 2025-06-19T21:22:11Z

No description provided.

bjsplat · 2025-06-19T21:23:58Z

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s).
To prevent this PR from going to the changelog marked it with the "skip changelog" label.

bjsplat · 2025-06-19T21:43:48Z

Snapshot stored with reference name:
refs/pull/16773/merge

Test environment:
https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/refs/pull/16773/merge/index.html

To test a playground add it to the URL, for example:

https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/refs/pull/16773/merge/index.html#WGZLGJ#4600

Links to test babylon tools with this snapshot:

https://playground.babylonjs.com/?snapshot=refs/pull/16773/merge
https://sandbox.babylonjs.com/?snapshot=refs/pull/16773/merge
https://gui.babylonjs.com/?snapshot=refs/pull/16773/merge
https://nme.babylonjs.com/?snapshot=refs/pull/16773/merge

To test the snapshot in the playground with a playground ID add it after the snapshot query string:

https://playground.babylonjs.com/?snapshot=refs/pull/16773/merge#BCU1XR#0

bjsplat · 2025-06-19T21:55:45Z

WebGL2 visualization test reporter:

https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/refs/pull/16773/merge/testResults/webgl2playwright/index.html

bjsplat · 2025-06-19T21:58:51Z

Visualization tests for WebGPU

https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/refs/pull/16773/merge/testResults/webgpuplaywright/index.html

packages/dev/core/src/Materials/PBR/pbrMaterial2.ts

bjsplat · 2025-06-20T21:08:51Z

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s).
To prevent this PR from going to the changelog marked it with the "skip changelog" label.

MiiBond · 2025-06-20T21:21:23Z

I've been playing with mixins to try to remove a lot of duplicate code from the materials. I figured, if we can do this, it'll make my life easier when working on PBR2. I started with using mixins for some defines like UV1, UV2, etc. and the image-processing defines. Then, I tried actually making the imageProcessing logic into a mixin. This removes a whole bunch of duplicated stuff.
See packages\dev\core\src\Materials\imageProcessing.ts
What do we think about this approach? It seems to work with typing but decorators don't work with anonymous class definitions so I had to apply them manually in the constructor.
If we're okay with this, I was thinking about trying to separate out some of the lighting logic (light map, reflection map, irradiance map, SH, etc.).
Ideally, it would be nice to have a bit more of a separation between the lighting and the material model.

bjsplat · 2025-06-20T21:25:23Z

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s).
To prevent this PR from going to the changelog marked it with the "skip changelog" label.

bjsplat · 2025-06-23T17:52:31Z

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s).
To prevent this PR from going to the changelog marked it with the "skip changelog" label.

MiiBond · 2025-06-23T17:53:49Z

packages/dev/core/src/Materials/imageProcessing.ts

+import type { ColorCurves } from "./colorCurves";
+
+// Explicit re-export of types to help TypeScript resolve them in declaration files
+export type { Observer } from "../Misc/observable";


I was getting errors in the npm run build:es6 involving ../.. import paths in this mixin.
An AI tool came up with the solution of adding these explicit exports. I don't really understand it but it works. It should be checked by someone who understands issues like this.

bjsplat · 2025-06-24T17:52:31Z

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s).
To prevent this PR from going to the changelog marked it with the "skip changelog" label.

MiiBond · 2025-06-24T19:50:16Z

I've noticed another issue with my changes that I don't understand. When running a Playground in Typescript mode, I get an error about not being able to find the BABYLON namespace. Any ideas what I might have done to cause that?

bjsplat · 2025-06-25T20:52:01Z

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s).
To prevent this PR from going to the changelog marked it with the "skip changelog" label.

bjsplat · 2025-06-26T20:24:52Z

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s).
To prevent this PR from going to the changelog marked it with the "skip changelog" label.

bjsplat · 2025-06-26T20:47:59Z

Building or testing the sandbox has failed.

If the tests failed, results can be found here:
https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/SANDBOX/refs/pull/16773/merge/testResults/

bjsplat · 2025-06-26T20:49:14Z

Building or testing the playground has failed.

If the tests failed, results can be found here:
https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/PLAYGROUND/refs/pull/16773/merge/testResults/

Popov72 · 2025-07-01T07:01:38Z

Just so I understand correctly, for now, you have simply copied and pasted the existing PBR vertex/fragment shaders to create the OpenPBR shaders?

I don't know if you have discussed the rewrite with @sebavan, but in my opinion, we should start from scratch without any defines and implement what is necessary for OpenPBR in the cleanest way possible. We can then configure the code with a few defines when necessary (but we should end up with far fewer definitions than we have now).

MiiBond · 2025-07-07T16:02:45Z

in my opinion, we should start from scratch without any defines and implement what is necessary for OpenPBR

This is actually what I started doing but I quickly realized that there is a ton of stuff that we still want to support like SH (in VS and FS), pre-filtered IBL, realtime IBL, analytic lights, etc. Also morph targets, skinning, etc. I'm still planning on removing ALL of the current material properties copied from PBRMaterial while retaining the logic for lighting, animation, etc. Ideally, we'll be able to separate lighting from material properties as much as possible in the shader...

bjsplat · 2025-07-10T01:34:57Z

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s).
To prevent this PR from going to the changelog marked it with the "skip changelog" label.

bjsplat · 2025-07-10T01:58:39Z

Building or testing the sandbox has failed.

If the tests failed, results can be found here:
https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/SANDBOX/refs/pull/16773/merge/testResults/

bjsplat · 2025-07-10T01:59:05Z

Building or testing the playground has failed.

If the tests failed, results can be found here:
https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/PLAYGROUND/refs/pull/16773/merge/testResults/

bjsplat · 2025-07-10T18:02:44Z

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s).
To prevent this PR from going to the changelog marked it with the "skip changelog" label.

bjsplat · 2025-07-10T18:26:39Z

Building or testing the sandbox has failed.

If the tests failed, results can be found here:
https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/SANDBOX/refs/pull/16773/merge/testResults/

bjsplat · 2025-07-10T18:27:22Z

Building or testing the playground has failed.

If the tests failed, results can be found here:
https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/PLAYGROUND/refs/pull/16773/merge/testResults/

bjsplat · 2025-07-10T19:36:48Z

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s).
To prevent this PR from going to the changelog marked it with the "skip changelog" label.

bjsplat · 2025-08-27T16:43:44Z

WebGL2 visualization test reporter:

https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/refs/pull/16773/merge/testResults/webgl2playwright/index.html

bjsplat · 2025-08-27T16:44:22Z

Visualization tests for WebGPU

https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/refs/pull/16773/merge/testResults/webgpuplaywright/index.html

bjsplat · 2025-08-27T17:26:46Z

WebGL2 visualization test reporter:

https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/refs/pull/16773/merge/testResults/webgl2playwright/index.html

bjsplat · 2025-08-27T17:27:12Z

Visualization tests for WebGPU

https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/refs/pull/16773/merge/testResults/webgpuplaywright/index.html

sebavan

looks all good to me, @Popov72 can you have a look at the PR ?

alexchuber

glTF exporter LGTM!

alexchuber · 2025-08-28T21:30:35Z

packages/dev/serializers/src/glTF/2.0/Extensions/KHR_materials_anisotropy.ts

+                    const geometryTangentTextureInfo = this._exporter._materialExporter.getTextureInfo(babylonMaterial.geometryTangentTexture);
+
+                    // TODO - if _useSpecularRoughnessAnisotropyFromTangentTexture isn't true, the anisotropy strength is in a separate
+                    // texture and needs to be merged into the blue channel of the aniosotropy texture for glTF. Is there a good way of


AFAIK, we don't yet have a clean, abstract way of composing channels from different textures into a new texture. If you were to do it right here and now, you'd have to do it all manually: read the original pixels, compose a new pixel array, decide on things like mimeType and sampling information, then finally export the glTF sampler, image, and texture info.

If you have some ideas in mind, feel free to implement them! But also don't hesitate to leave a TODO or note-- I'd be happy to fill this in (and other similar cases, like the metal/rough textures in _convertMetalRoughFactorsToMetallicRoughnessAsync) later.

bghgary

Some high-level comments for now. I will be discussing with Seb on how to not branch so much in the loader.

bghgary · 2025-08-28T21:03:27Z

packages/dev/loaders/src/glTF/2.0/glTFLoaderExtension.ts

     * @returns A promise that resolves when the load is complete or null if not handled
     */
-    loadMaterialPropertiesAsync?(context: string, material: IMaterial, babylonMaterial: Material): Nullable<Promise<void>>;


This shouldn't be necessary. The loader extensions already have access to the main loader and the parent loader, so they already have access to this flag.

bghgary · 2025-08-28T21:05:55Z

packages/dev/loaders/src/glTF/2.0/glTFLoader.ts

@@ -91,6 +92,8 @@ const LazyLoaderAnimationModulePromise = new Lazy(() => import("./glTFLoaderAnim

 export { GLTFFileLoader };

+let PBRMaterialClass: typeof PBRMaterial | typeof OpenPBRMaterial;


I don't think this will work correctly as a global. Since the determination of what class to use is per loader, this variable should be stored in the loader class itself, or it may get overridden when two assets are loaded simultaneously. Extensions can also access this property to avoid duplicating the logic.

bghgary · 2025-08-28T21:08:25Z

packages/dev/loaders/src/glTF/2.0/glTFLoader.ts

+                }
+                mat.metallic = properties.metallicFactor == undefined ? 1 : properties.metallicFactor;
+                mat.roughness = properties.roughnessFactor == undefined ? 1 : properties.roughnessFactor;
+                babylonMaterial = mat;


This line is unnecessary.

bghgary · 2025-08-28T21:31:44Z

packages/dev/loaders/src/glTF/2.0/glTFLoader.ts

+                            (babylonMaterial as OpenPBRMaterial)._useRoughnessFromMetallicTextureGreen = true;
+                            (babylonMaterial as OpenPBRMaterial)._useMetallicFromMetallicTextureBlue = true;


The glTF loader tries to make sure all shader variants are accounted for in the first synchronous pass. I'm guessing these flags will cause the shader to be different. If so, these should be moved to below, like with the PBRMaterial flags. Same goes for any flags that may affect the shader. We need to make sure they are set synchronously and not in a promise continuation.

bghgary · 2025-08-28T22:35:37Z

@MiiBond Here is my idea for avoiding so much branching in the loader.

If we can add a class like this:

class OpenPBRMaterialAdapter {
    private _oldMaterial: PBRMaterial

    public constructor(oldMaterial: PBRMaterial) {
        this._oldMaterial = oldMaterial;
    }

    public set baseMetalness(value) { this._oldMaterial.metallic = value; }
}

Then, we can use this adapter when openPbr is not being used but still have the code be like it's using the open pbr one. Let me know if this doesn't make sense.

deltakosh reviewed Jun 20, 2025

View reviewed changes

packages/dev/core/src/Materials/PBR/pbrMaterial2.ts Outdated Show resolved Hide resolved

deltakosh requested review from Popov72 and sebavan June 20, 2025 15:33

MiiBond commented Jun 23, 2025

View reviewed changes

deltakosh changed the title ~~Added PBR2 material and option to load using glTF~~ Added OpenPBR material and option to load using glTF Jun 24, 2025

MiiBond force-pushed the mbond/pbr2-test-wip branch from fe87eca to 3a5b609 Compare June 25, 2025 20:37

MiiBond force-pushed the mbond/pbr2-test-wip branch from df6a640 to e908eee Compare July 10, 2025 01:33

Mike Bond and others added 19 commits August 27, 2025 09:01

Remove unused vis test image

ec2cce9

Fix for UMD build tests

8d59e43

Split OpenPBR metal and rough textures

6d94bb4

WIP adding anisotropy params

e800ddb

Fix uniform names for OpenPBR when not using UBO

fb6be12

More OpenPBR fixes for non-UBO

37ef9d9

possible fix

2899e88

possible fix

aa94392

possible fix

69e5823

Revert baseParticleSystem

3a9f6b7

WIP - OpenPBR anisotropy

1ba1430

Adding asset with tangents for aniso tests

c6366dc

WIP OpenPBR anisotropy export

0d5918f

Update OpenPBR aniso reference images

4ea3f9a

Update OpenPBR aniso vis tests

abe9516

Update vis reference for anisotropy

03bee59

WIP OpenPBR anisotropy for IBL

41ca3f0

Fix import

4aacaf1

Fix import issue

294a1e5

MiiBond force-pushed the mbond/pbr2-test-wip branch from 350cb2f to 294a1e5 Compare August 27, 2025 16:12

Fix import issue

a597577

sebavan reviewed Aug 28, 2025

View reviewed changes

alexchuber reviewed Aug 28, 2025

View reviewed changes

bghgary reviewed Aug 28, 2025

View reviewed changes

		@@ -91,6 +92,8 @@ const LazyLoaderAnimationModulePromise = new Lazy(() => import("./glTFLoaderAnim

		export { GLTFFileLoader };

		let PBRMaterialClass: typeof PBRMaterial \| typeof OpenPBRMaterial;

		(babylonMaterial as OpenPBRMaterial)._useRoughnessFromMetallicTextureGreen = true;
		(babylonMaterial as OpenPBRMaterial)._useMetallicFromMetallicTextureBlue = true;

OpenPBRMaterial (including loading and exporting glTF) #16773

Are you sure you want to change the base?

OpenPBRMaterial (including loading and exporting glTF) #16773

Conversation

MiiBond commented Jun 19, 2025

Uh oh!

bjsplat commented Jun 19, 2025

Uh oh!

bjsplat commented Jun 19, 2025

Uh oh!

bjsplat commented Jun 19, 2025

Uh oh!

bjsplat commented Jun 19, 2025

Uh oh!

Uh oh!

bjsplat commented Jun 20, 2025

Uh oh!

MiiBond commented Jun 20, 2025

Uh oh!

bjsplat commented Jun 20, 2025

Uh oh!

bjsplat commented Jun 23, 2025

Uh oh!

MiiBond Jun 23, 2025

Choose a reason for hiding this comment

Uh oh!

bjsplat commented Jun 24, 2025

Uh oh!

MiiBond commented Jun 24, 2025

Uh oh!

bjsplat commented Jun 25, 2025

Uh oh!

bjsplat commented Jun 26, 2025

Uh oh!

bjsplat commented Jun 26, 2025

Uh oh!

bjsplat commented Jun 26, 2025

Uh oh!

Popov72 commented Jul 1, 2025

Uh oh!

MiiBond commented Jul 7, 2025

Uh oh!

bjsplat commented Jul 10, 2025

Uh oh!

bjsplat commented Jul 10, 2025

Uh oh!

bjsplat commented Jul 10, 2025

Uh oh!

bjsplat commented Jul 10, 2025

Uh oh!

bjsplat commented Jul 10, 2025

Uh oh!

bjsplat commented Jul 10, 2025

Uh oh!

bjsplat commented Jul 10, 2025

Uh oh!

bjsplat commented Aug 27, 2025

Uh oh!

bjsplat commented Aug 27, 2025

Uh oh!

bjsplat commented Aug 27, 2025

Uh oh!

bjsplat commented Aug 27, 2025

Uh oh!

sebavan left a comment

Choose a reason for hiding this comment

Uh oh!

alexchuber left a comment

Choose a reason for hiding this comment

Uh oh!

alexchuber Aug 28, 2025

Choose a reason for hiding this comment

Uh oh!

bghgary left a comment

Choose a reason for hiding this comment

Uh oh!

bghgary Aug 28, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

bghgary Aug 28, 2025 •

edited

Loading

bghgary Aug 28, 2025 •

edited

Loading